0
Review Article

Misalignment Compensation for Full Human-Exoskeleton Kinematic Compatibility: State of the Art and Evaluation OPEN ACCESS

[+] Author and Article Information
Matthias B. Näf

Mechanical Engineering Department,
Vrije Universiteit Brussel,
Pleinlaan 2,
Brussels 1050, Belgium
e-mail: Matthias.Basil.Naf@vub.be

Karen Junius

Mechanical Engineering Department,
Vrije Universiteit Brussel,
Pleinlaan 2,
Brussels 1050, Belgium
e-mail: Karen.Junius@vub.be

Marco Rossini

Mechanical Engineering Department,
Vrije Universiteit Brussel,
Pleinlaan 2,
Brussels 1050, Belgium
e-mail: Marco.Rossini@vub.be

Carlos Rodriguez-Guerrero

Mechanical Engineering Department,
Vrije Universiteit Brussel,
Pleinlaan 2,
Brussels 1050, Belgium
e-mail: Carlos.Rodriguez.Guerrero@vub.be

Bram Vanderborght

Professor
Mechanical Engineering Department,
Vrije Universiteit Brussel,
Pleinlaan 2,
Brussels 1050, Belgium

Dirk Lefeber

Professor
Mechanical Engineering Department,
Vrije Universiteit Brussel,
Pleinlaan 2,
Brussels 1050, Belgium

1Corresponding author.

2M. B. Näf and K. Junius contributed equally to this work.

Manuscript received October 19, 2017; final manuscript received November 19, 2018; published online February 13, 2019. Assoc. Editor: Thao (Vicky) Nguyen.

Appl. Mech. Rev 70(5), 050802 (Feb 13, 2019) (19 pages) Paper No: AMR-17-1076; doi: 10.1115/1.4042523 History: Received October 19, 2017; Revised November 19, 2018

The use of exoskeletons by the elderly, disabled people, heavy labor workers, and soldiers can have great social and economic benefit. However, limitations in usability are impeding the widespread adoption of exoskeletal devices. Kinematic compatibility, comfort, volume, mass, simplicity, expandability, and the ability to transmit forces, relative angles between the exoskeleton and the human, and the donning and doffing procedure need to be considered. Over the last decades, a large number of exoskeletons have been developed, to assert kinematic compatibility and compensate for misalignment. To such a degree, that it has become difficult to keep an overview of the different strategies. Therefore, this review article presents an extensive overview of different misalignment compensation strategies existing in the literature. Further, these strategies are organized in nine categories, evaluated and discussed around the exoskeleton's application domain and its specific requirements and needs.

Interest in exoskeleton technology is increasing with strong momentum worldwide. An exoskeleton is an artificial device that provides physical assistance or power augmentation to enhance human physiological capabilities. Power augmentation exoskeletons have been constructed for a wide range of application domains including military (Berkeley Lower Extremity Exoskeleton (BLEEX) [1]), industrial workers [2,3], and healthcare [4] applications, to state a few.

As stated in Ref. [5], “While assistive devices can have a profound effect on a person's abilities, such devices have a high abandonment rate” with approximately one-third of all assistive devices abandoned according to Scherer [6]. Mobility aids are more frequently abandoned than other categories of devices [7].

For exoskeletons in particular, limitations in usability are impeding the widespread adoption of such devices in daily life [810]. In order to improve the mobility of the wearer, exoskeletons should be able to provide sufficient support to induce a reduction of user effort to manageable levels [11], thus allowing them to move around for longer periods of time. Because user exertion and onset of fatigue are strongly correlated to the energy consumed during motion [12,13], an increase in energy consumption while wearing an exoskeleton should at all times be avoided [14].

Ideally, an exoskeleton reduces energy consumption of the user during movement. This goal is shared mostly by lower limb exoskeletons [15,16]. Despite this common goal, only a limited number of devices have successfully reduced the energy consumption of their users [12,13,1724].

The difficulty in reducing the metabolic cost lies in the fact there is no general consensus on how to achieve this more effectively [25]. Mass and inertia of the device are considered to be important influential factors [8,26,27], as well as the disturbance of the wearer's natural motion patterns [2729]. As for the actuation strategy, Galle [29] stated that interaction of an exoskeleton with the human system can only be successful when assistance comes at the right moment, with just enough power, for just the necessary joints.

This greatly complicates the matter as there is no such thing as the average person [30,31], just as there is no typical disability [5]. Consequently, there is no such thing as a “one-strategy-fits-all”-actuation principle and therefore, the design of an actuation system is only beneficial for a very limited group of target subjects. Moreover, the design of the underlying kinematic structure, that allows the wearer to move through the surrounding environment without feeling hindered, is not dependent on the deficiencies or capabilities of a particular user and is thus advantageous for all users of exoskeletal devices. In order to achieve this, a crucial step is to ensure that the kinematic structure of the exoskeleton is compatible with the human joints it assists [32].

Kinematic Compatibility.

Several strategies can be used to ensure kinematic compatibility (see Fig. 1). The two main strategies are: mimicking the anatomical joint, or using a simplified exoskeleton joint structure to ensure proper alignment between the axes of rotation.

In anthropomorphic exoskeletons, there is typically a one-to-one mapping between the degrees-of-freedom (DOFs) of the device and its wearer, i.e., each rotational DOF of the exoskeletons corresponds to a single DOF of the human wearing it. Considering this definition, the exoskeletons with the simplified joint structure can be further divided, depending on the nature of the exoskeleton device that is used: anthropomorphic devices versus nonanthropomorphic devices and end effectors.

Fully mimicking the kinematics of the anatomical joint requires exact knowledge of the joint's instantaneous axes of rotation at any given time during motion. Specifically, for the knee joint, efforts have been undertaken to mimic the underlying kinematics of the human joint. A concept was developed by Bertomeu et al. [33] and further scaled into a hardware device by Tucker et al. [34]. However, an in-depth treatment of joint mimicking devices is outside the scope of this paper, especially considering that human joints are covered with a multitude of tissues. Therefore, it is difficult to determine the exact location of the joint rotation axes in the absence of imaging devices [35]. A more feasible alternative is to estimate the location of the joint center. Unfortunately, this method has shown poor accuracy [3638].

Additionally, substantial inter- and intrapatient differences further decrease the accuracy of the estimation methods [39,40], making it impossible to obtain an exact model of anatomical joint kinematics. Therefore, no attempt to perfectly mimic the joint kinematics can ever achieve a flawless compatibility between human and exoskeleton. In order to achieve compatible kinematics with high-quality interactions, a different approach must be pursued [41].

It is a common practice to simplify the human joints in exoskeleton design [42,43]. By only considering the dominant rotations [43], the human joints are commonly modeled as a set of 1DOF hinges [44,45].

In anthropomorphic exoskeletons, where any hinge of the device directly corresponds to a DOF of the human limb, correct alignment of the axes of rotation is crucial to ensure the correct transfer of torque from the exoskeleton to the wearer [35,46]. If these are not properly aligned, parasitic forces and torques are induced, which may cause discomfort or pain and may potentially even lead to long term injury, or dislocation of the joint by frequent use of the device [39,41,4749].

The use of nonanthropomorphic exoskeletons or end-effector devices removes the burden of ensuring proper alignment, because their hinges do not directly correspond to the human DOFs. These devices only match the global behavior of the human limbs and, therefore, alignment is not an issue [42,50]. On the other hand, they are prone to other difficulties. In rehabilitation applications, it is crucial to have individual control over at least some of the included DOFs in order to ensure targeted interventions and reproducible motion patterns [40,48]. It is also important that nonimpaired joints are unaffected by the robot [40]. Neither can be assured by use of end effectors and nonanthropomorphic devices [8,39].

Additionally, end-effector robots cannot resolve the natural human limb redundancy and therefore have the potential to apply harmful loading to the human joints [51]. For all the reasons mentioned above, the use of anthropomorphic exoskeletons is preferred in assistive applications and rehabilitation. It is clear then that additional effort is required to address the issue of misalignment of the rotation axes in order to ensure complete and safe kinematic compatibility.

In this manuscript, an extensive review of misalignment compensation strategies is presented. These strategies are used mainly to improve usability in the exoskeleton's application domain. In Sec. 2, the motivation, classification, and requirements for these mechanisms are presented. In Sec. 3, the exoskeleton literature and their compensation strategies are classified and evaluated with respect to the previously defined requirements. The results obtained from this review are presented in Sec. 4 and discussed in Sec. 5. The manuscript finishes with Sec. 6 where conclusions are drawn.

In the following, first the motivation for misalignment compensating mechanisms is presented, followed by a classification of the different mechanisms. Finally, requirements for a mobile exoskeleton are defined based on an extensive literature review with misalignment compensating mechanisms.

Motivation.

Assuming an initial displacement between a human joint and an exoskeleton, if both are displaced by an angle α (See Fig. 2), the initial displacement results either in a radial displacement in combination with a tangential displacement (See Fig. 2(a)), or a radial displacement in combination with a small angular displacement (See Fig. 2(b)) if we assume that there is a brace [35].

In reality, a brace prevents free displacement. If the physical connection between the robot and the human were perfectly rigid, kinematic incompatibility between the two systems would prevent motion [52]. In fact, the rotation of an exoskeleton joint, which is misaligned with respect to a human joint, is only possible due to the presence of soft tissues and cartilages that can withstand large deformations [52].

Despite these deformations, parasitic forces are still created between the brace and the skin. Peak forces and parasitic torques of up to 230 N and 1.5 N·m, respectively, have been documented in absence of any actuation [53]. These can cause discomfort or pain and may potentially even lead to long-term injury, dislocation of the joint by frequent use of the device [39,41,4749]. Additionally, the kinematic mismatch between the user and the device also influences the voluntary range of motion (ROM) [54], natural patterns of movement [55,56], and muscle activation patterns [57].

The total misalignment between an exoskeleton and its user is usually a combination of four separate effects [8,58]

  1. (1)Kinematic mismatch
  2. (2)Migration of the instantaneous center of rotation
  3. (3)Initial offset
  4. (4)Movement mismatch

Kinematic mismatch is related to the number of DOFs of the human joint. If the exoskeleton does not include all the DOFs that are present in the human joint, misalignment is inevitable. This phenomenon is more obvious for the human shoulder. Although it is most often modeled as a spherical joint, the human shoulder also has two translational DOFs, allowing it to move up and down and from the front to the back [59,60]. Because a spherical exoskeleton joint cannot account for these movements, a misalignment between exoskeleton and wearer always exist.

Migration of the instantaneous center of rotation is connected to the fact that exoskeleton joints are frequently designed, by neglecting the small translations seen in its human counterpart [43] (see Fig. 2(a)). For example, the human knee is most often simplified as a one DOF hinge joint. Yet, due to a rolling motion between the femur and the tibia, its instantaneous center of rotation moves during flexion [61]. Due to simplifications of the joint, the exoskeleton cannot account for this migration during movement. This induces misalignment between the rotation axes.

Initial offset concerns the circumstance, that it is never possible to determine the exact position of the joint axes, because the human bone structure is covered with muscles and tissue. Therefore, in an attempt to align the rotation axes, an initial offset of several centimeters is to be expected [62,63]. Additionally, because misalignment induces soft tissue deformation and slippage during motion, the mismatch between the rotation axes is likely to increase with movement. A mismatch of up to 10 cm is not considered out of the ordinary, even if the initial alignment was satisfactory and the braces were tightly fixed [53,54].

Movement mismatch accounts for the fact that the interfaces of an exoskeleton can migrate. This effect might negatively affect the kinematic compatibility.

Classification.

Existing anthropomorphic exoskeletons (see Fig. 1) require misalignment compensation techniques to guarantee high quality interactions. For clarity sake, we need to start by discussing kinematic redundancy techniques. When the kinematic redundancy technique is used, kinematic compatibility is ensured by adding additional DOFs to the device. Typically, three additional degrees-of-freedom are introduced (See Fig. 3). These additional DOFs consist of a combination of revolute and prismatic joints. Depending on the type of the additional elements, a division can be made in three categories: RRP, RPP, and RRR, where “R” stands for a revolute joint and “P” stands for a prismatic joint. Kinematic redundancy techniques that do not fit in any of the three categories are grouped in the “other” category. Each of the different groups is extensively discussed in Sec. 3.

Misalignment compensation techniques can be implemented in different ways and will serve as main groups for the rest of the classification: manual alignment, use of compliant elements, and addition of kinematic redundancy. Two of these categories can be further subdivided. When compliant elements are used, they can either be introduced at brace level, i.e., compliant braces or in the frame. Additionally, compliant exoskeleton joints can be also implemented.

We refer the reader to Fig. 4 for a synthesis of the classification of misalignment compensation techniques.

Requirements.

The technical requirements of assistive kinematically compatible exoskeletons arise from three different sources. First of all, kinematic compatibility entails that the user of an exoskeleton can move unhindered while wearing the device. This calls for a specific set of requirements to allow this feature. Second, because of the assistive application, several requirements are introduced that are related to the metabolic benefit of the device. An assistive exoskeleton can only improve the mobility of its wearer if it does not have a negative impact on his or her metabolic cost. Finally, the rate of abandonment of mobility aids is the highest all assistive devices. Therefore, several technical requirements are introduced that have a track record of improving device adoption. Each of these requirements is discussed further in Secs. 2.3.12.3.3.

Requirements: Kinematic Compatibility.

The exact kinematics of the human joints cannot be reproduced with exoskeleton joints if they consist of simple kinematic pair with 1DoF. For three-dimensional joints, this entails the presence of three perpendicular rotation axes: first for flexion/extension, a second for ab-/adduction, and a third for internal/external rotation. Because perfect alignment of the exoskeleton's hinge joints with the human axes of rotation is unattainable, all three hinge joints need to be equipped with additional elements to compensate for the effects of misalignment with the human joints. With all these elements included, the wearer of any exoskeleton should be able to maintain full ROM.

Requirements: Metabolic Benefit.

Although the design of a fully compatible exoskeleton joint is a positive accomplishment, the actual goal of introducing kinematic compatibility is to improve the metabolic performance of a powered exoskeleton. With this goal in mind, it is important that the hip exoskeleton is as light and compact as possible in order to reduce its (negative) impact on the metabolic consumption of the user [64,65].

Note that assistive exoskeletons need to be equipped with an actuating system, providing its wearer with assistance during task execution. In order to allow this, the kinematic structure needs to be capable of transmitting a torque from the actuator to the human. Because the targeted use of the exoskeleton is assisting users during daily life, there is no need to provide the full biological torque. The intention is to keep the user actively participating during task execution to discourage muscle degeneration due to disuse.

Requirements: User Acceptance.

User acceptance is an important criterion that is often forgotten in the design process. Indeed, many assistive devices that can improve a person's abilities and quality of life are usually abandoned [7]. Given the influence that a user's perception has on the adoption of an assistive device, the authors want to stress the inclusion of user-acceptance criteria in the evaluation of compensation strategies. This, indeed, is significantly easier to provide an initial design that meets such requirements than it is to adapt an existing prototype to meet the requirements of user acceptance [66].

A significant factor associated with continued or discontinued use of technology is the relative advantage of the device [67]. Important criteria that users utilize to assess the relative advantage are: effectiveness, durability, safety, and comfort [68,69].

Effectiveness in this case is directly related to the amount of hindrance that the wearer experiences when moving while wearing the exoskeleton and its effect of the metabolic cost. Effective operation is already ensured by the adoption of the requirements that were given earlier, ensuring kinematic compatibility and metabolic benefit.

Durability requires the use of robust components that do not need regular maintenance.

Safety requires that the device does not pose a risk for its user during operation.

Comfort is a particularly important element because it can lead to disuse no matter the performance in other areas: users are not likely to use a device that performs well but chafes the skin during operation, or is obtrusive during activities of daily living.

Other factors that have strong influences on the adoption by users are the physical appearance and ease of use of the assistive devices [5,70]: they must be esthetically pleasing, socially acceptable, portable, lightweight, and easy to use independently in order to qualify for user acceptance. Because what is esthetically pleasing and socially acceptable are subjective criteria that are highly dependent on the specific target groups. This consideration will not be included in the evaluation in Sec. 4. On the other hand, portability, mass, and ease of use are objective parameters that can be easily assessed. Apart from minimizing the mass, portability can be improved, by reducing the volume of the exoskeleton and keeping it as close to the human body as possible. Ease of use translates into the possibility of autonomously donning the device, without the need for multiple adjustments. It is important to stress that kinematic compatibility can greatly reduce the complexity of adjusting the exoskeleton to the wearer.

In the following, exoskeletons from the literature are classified according to their misalignment concept (see Fig. 4).

Manual Alignment.

The vast majority of the state-of-the-art devices employ the strategy of manual alignment. Due to the variability of joint location between individuals, this task requires precise anthropomorphic measurements of a person, in order to determine the exact joint location [58]. Furthermore, the exoskeleton system has to be adjusted so that the axes of the device coincide with those of the human. This is often a cumbersome operation that requires skill and knowledge of human anatomy.

Typically, an exoskeleton is connected to the human with braces around the limbs. Its position is therefore variable with respect to the skeletal system due to deformations of the soft tissue [40]. Additionally, the braces of an exoskeleton are prone to slippage during operation [50,7173]. Offsets of as much as 10 cm between exoskeleton and users' axes of rotation have been recorded even with proper initial alignment [53]. It is thus safe to conclude that perfect manual alignment is unattainable.

Nevertheless, this approach can give good results, provided the technician is skilled [74]. Because manual alignment only offers a solution for initial offset misalignment, the result can only be good, if the human DOF closely approximates a hinge joint.

Additionally, the process can be time-consuming: fitting times of 10–20 min are not out of the ordinary [75,76]. Some examples of exoskeletons in which this strategy is used are: the Lokomat [77,78], HAL [79,80], Ekso [81], ReWalk [82], CORBYS [83], and Indego [14]. All of these include hip and knee flexion/extension joints for both legs.

Compliant Elements.

A second way to deal with joint misalignment is the use of compliant elements. These can be implemented either at the brace, frame, or joint level. Note that the compliance introduced in the kinematic structure, i.e., in braces, the frame or the joints, is independent of the actuation strategy.

Compliant Elements: Frame and Brace Level.

In devices with manually aligned joints, the exoskeleton is usually connected to the human by semirigid braces. Misalignment effects due to kinematic mismatch and migration of the instantaneous center of rotation can be reduced by deformation of the braces [52]. This has mostly been employed in rehabilitation devices and has already been documented for Lokomat and Rewalk [74].

Another possibility is to incorporate compliance in the frame of the exoskeleton. In Ref. [84], the internal/external rotation DOF of a hip exoskeleton was implemented by mounting two sliding structures onto a flexible, sheet metal waist band (see Fig. 5). Deformation of the band can account for differences in waist size and curvature among subjects. No perfect initial alignment is required in this case.

Additionally, the authors in Ref. [85] use flexible frame structures (see Fig. 6). These structures were designed, to be rigid in the direction where torque needs to be transmitted to the user and flexible in other directions. Therefore, they can be tightly fitted to each individual wearer. Slight deformations are used to alleviate misalignment effects outside of the plane, where torque is applied. However, this structure cannot compensate for misalignment in the flexion/extension rotation axis [85].

Because frame or brace-level compliance can only compensate for small amounts of misalignment [32], its effectiveness is still highly dependent on the donning procedure. Therefore, the closer the initial alignment of rotation axes and the closer the human joint approximates the exoskeleton hinge joint, the better the results will be.

Compliant Elements: Joint Level.

Compliant elements can also be implemented at the joint level. The joint axis is then replaced by a compliant coupling, which gains its degrees-of-freedom from an elastic deformation [86].

Typically, traditional joints offer low stiffness around their rotation axis and a high stiffness in any other direction. This is not the case for compliant joints [87]. The stiffness in any direction depends on the structural properties of the joint. A review of several compliant joint designs has been offered in Ref. [87] and some examples are shown in Fig. 7.

The advantage of a lowered stiffness around an axis other than the principal axis is that this allows for small compensatory movements in case of a misalignment.

A compliant joint element based on leaf springs has been implemented in an elbow exoskeleton by Kleinjan et al. [86]. It allows for a compact design and auto-aligns itself with the human axis of rotation. The entire device fits underneath the clothing of a user. A downside of these elements is that, due to their nature, movement of the joint is always accompanied with a torque around the joint axis. If zero-torque operation is required, an additional actuator must be introduced to counteract the passive torque arising from joint flexion. On the other hand, a non-zero torque can be used to counteract the mass of the device, as noted in Ref. [86].

In the quasi-passive exoskeleton of Walsh et al. [28], ankle inversion/eversion is accounted for deforming the foot plate that is attached to the wearer's shoes. As such, the foot plate operates as a compliant joint.

Adding Kinematic Redundancy.

A further possibility to introduce kinematic redundancy in an exoskeleton is to add more DOFs. This approach has been getting more attention over the last decade and is largely founded in the research performed by Schiele and Visentin [88] for the European Space Agency (ESA). Later, Stienen et al. [39,89] built on this knowledge to develop an arm exoskeleton for rehabilitation. Their goal was to match the kinematics of the human shoulder complex and eliminate the detrimental effects of joint axis misalignment. The work of the latter group is mostly responsible for increasing the visibility of misalignment compensation in exoskeleton applications.

Revolute-Revolute-Prismatic Joint Configuration.

In the revolute-revolute-prismatic (RRP) joint configuration, an original single-DOF hinge joint (see Fig. 8(a)) is outfitted with an additional revolute joint parallel to the original one and an additional prismatic joint perpendicular to the original hinge joint axis (see Figs. 8(b) and 8(c)).

This configuration was first implemented in an elbow module of the ESA exoskeleton, which allows joint flexion and extension [35], even if the exoskeleton and the human joint are misaligned.

In Ref. [54], it was shown that an exoskeleton that features passive compensation joints as described above can lower interaction forces by up to 70% and torques by up to 60%. Additionally, passive compensation joints can restore natural ROM if sufficient stroke margins are foreseen in the design.

Näf et al. [90] used an RRPR mechanism to compensate for misalignment between the axes of rotation of the human lumbo-sacral (L5/S1) joint(s) and the exoskeleton joint(s) (see Fig. 9) in their back support exoskeleton [90]. In their design, the first two rotation axes are implemented with the elasticity of bending beams.

D'Elia et al. [91] presented a hip exoskeleton (Fig. 10), which includes an ab-/adduction and an internal/external hinge on top of the conventional flexion/extension DOF. Several additional passive DOFs are present in the prototype as well (Fig. 10). Rotational and prismatic joints at brace level could account for ab-/adduction misalignment, implementing an RRP mechanism.

Cai et al. [92] presented a self-adjusting knee exoskeleton (see Fig. 11(a)). The authors mention that the position of the joint with respect to the human knee can be changed by using a vertical slider. This leads us to assume that the flexion/extension hinge joint is manually aligned. The paper is unclear on whether the slider is locked during operation, after manual adaptation.

In addition to three perpendicular hinges and the slider, the device also includes a fourth hinge, allowing rotation in the frontal plane, and a fifth one allowing rotation in the transverse plane. These latter hinges were, reportedly, added to compensate for the offset between the devices rotation and abduction axes and those of the human knee. If the vertical slider in the design of Cai remains free during operation, misalignment of the abduction axis is compensated by its movement accompanied with the additional abduction joint as first presented by Schiele and Van Der Helm [35]. For the internal/external rotation, the slider is oriented in the wrong direction, the required linear movement should be allowed by a limited amount of deformation of the soft tissue. The inclusion of the passive elements has led to a bulky prototype. In a more recent version of the exoskeleton [93] (Fig. 11(a)), the number of sliders in thigh and shank elements has been increased and the authors do mention that they remain free during operation. In the newer version of the exoskeleton, the additional passive elements, i.e., sliders and revolute joints, are housed in between the exoskeleton frame and the braces connecting the device to the human's thigh and shank. At each location, two perpendicular sliders are included in the sagittal plane: one along the limb segment and another perpendicular to the limb segment. Due to the addition of the perpendicular slider, rotation axis misalignment can now be compensated for by the RRP mechanism.

Revolute-Prismatic-Prismatic Joint Configuration.

The RPP architecture (see Figs. 11(b) and 12) was first proposed as a misalignment compensation mechanism by Stienen et al. [89] in a rehabilitation arm exoskeleton that captures the complex kinematics of the human shoulder (see Fig. 12).

Apart from three rotational DOFs, i.e., flexion/extension, ab-/adduction, and internal/external rotation, the human shoulder complex also displays two distinct translational motions: an up-down movement, called elevation and depression, respectively, and a forward-backward movement, called protraction and retraction, respectively [60]. These are not represented in the ball-and-socket model that is often used to model the shoulder [59]. Therefore, researchers have searched for ways to implement these DOFs in their rehabilitation device prototypes.

Nef et al. [59] presented the rehabilitation robot ARMin. This device implements only the three rotational DOFs of the shoulder. These are all actuated. During experiments with ARMin, subjects complained of being forced into uncomfortable postures during elevation of the arm [94]. To account for this, two passively coupled translational DOFs were added in the design of the ARMin II [94] (see Fig. 13(a)). These DOFs account for the elevation of the shoulder rotation axis and are mechanically coupled to the elevation of the arm. Experimental data showed that the elevation of the exoskeleton was a good match for the elevation of the anatomical shoulder joint. However, the mechanical coupling was deemed to be too large, too heavy, and added a significant amount of inertia to the system [42].

Park et al. [95] equipped their IntelliArm with linear sliders to incorporate all shoulder DOFs, including a translational DOF to accommodate for side to side trunk movements (see Fig. 13(b)). Shoulder flexion/extension, ab-/adduction, rotation, and elevation/depression are actively assisted by rotational actuators for the first three DOFs and a vertical, linear actuator for the latter DOF. Shoulder retraction/protraction and side-to-side trunk movements are accommodated by two passive perpendicular sliders in a plane (X–Y) parallel to the ground.

Stienen et al. [39] used a similar kinematic design for the shoulder, where three perpendicular sliding mechanisms are followed by three rotational actuators in the Dampace exoskeleton. In contrast to the ARMin and IntelliArm devices, the linear and rotational DOFs of the Dampace exoskeleton provide a controllable resistive braking action, rather than assistive actuation. Apart from their use in accommodating shoulder elevation and retraction, Stienen et al. [39] stated that the linear DOFs also allow the robot axes to self-align with the anatomical axes of rotation. The principle was recognized by studying the work of Schiele and Visentin on the ESA exoskeleton [88] and is clarified in Fig. 12.

Because of the linear linkage mechanism, the Dampace exoskeleton was found to be too heavy and subject to high friction forces. The required linkage movements were high; therefore, an alternative approach was used in the successor: LIMPACT [39]. Instead of a linear guidance system, the LIMPACT uses a system with revolute joints in a parallelogram structure. Three parallelograms allow for translation in the x, y, and z directions, equivalent to the linkage system in Dampace [96]. It follows that LIMPACT is also equipped to alleviate the interaction forces caused by misalignment.

In the recent exoskeleton of Cai et al. [93] (see Fig. 11(a)), which was already mentioned in Sec. 3.3.1, several passive elements are housed in between the exoskeleton frame and the braces connecting the device to the human's thigh and shank. Misalignment of the knee abduction/adduction and rotation axis could be compensated for by following the RRP strategy. Due to the incorporation of the additional sliding elements, misalignment of the flexion/extension axis is also taken care of since the exoskeleton joint can move with respect to the human joint as seen in the RPP setup. Although not specified in the paper [93], it is important to note that the two sliding elements in the thigh segment could also account for misalignment compensation of the hip flexion/extension axis. In this case, the R element is included in the spherical hip joint, and both sliders fulfil the PP function.

Bartenbach et al. [11] utilized the RPP setup for knee misalignment compensation of the flexion/extension axis in their hip-knee exoskeleton. The knee flexion/extension joint is modeled as a simple hinge, yet the thigh and shank brace are allowed to slide along the exoskeleton frame by the introduction of two linear DOFs (see Fig. 11(a)). The main advantage of adding passive elements in the connection to the braces rather than in the exoskeleton frame itself is that the mass of the device can still be transferred to the ground through the joint structure. The downside of this strategy is that when the knee joint is fully extended, the mechanism only allows motion between the exoskeleton hinge and knee joint along the vertical axis of the frame. Therefore, any horizontal misalignment that is present cannot be accounted for. Additionally, any sliding motion of the exoskeleton frame with respect to the human would result in a displacement of not only the knee hinge but also of the hip hinge. Because it is unlikely that both joints require the same corrective motion in order to compensate for misalignment, a more beneficial position for one joint would likely result in a disadvantageous position for the other.

Lee et al. [84,85] incorporated all three hip movements into the hip of their S-assist exoskeleton. Besides being used for adaptation purposes, the vertical sliding mechanism (Fig. 5) has the additional function of compensating the movement between the device and the human due to misalignment of the ab-/adduction axis.

Revolute-Revolute-Revolute Joint Configuration.

The RRR architecture was first used in the wrist module of the ESA exoskeleton [88]. Here, the wrist ab-/adduction mechanism consists of a set of three parallel rotational DOFs connected by two additional linkages (Fig. 14(a)). By driving joint 2, ab-/adduction is enforced, causing passive motion of joints 1 and 3 as a result of the misalignment of rotational axes. The same mechanism has later been implemented into the elbow joint of the Dampace exoskeleton [39,89] and an expansion module for the wrist [97]. The kinematic structure and operating principle of the elbow joint in the Dampace exoskeleton are similar to the one implemented by Schiele et al. for the wrist. To explain its functionality, we can refer to Fig. 14(a). The three joints are equipped with drums, which are connected by a cable drive [89]. The drum on the upper arm is actuated (joint 1; TAct) and a torque applied here is transmitted through the cable until joint 3, rotating the lower arm. Adding the two additional DOFs allows for translation of the joint to be independent of rotation. Therefore, the interaction forces are removed by a movement of the three rotation axes with respect to each other.

The wrist module of Beekhuis et al. [97] is shown in Fig. 14(b). The mechanism has been practically implemented as a series of rotational DOFs and rigid linkages that have been mounted in a double parallelogram setup, with a one-to-one relation between rotation of the motor and rotation of the wrist. Therefore, the orientation of the wrist can be directly measured by a read-out of the motor encoders. A visualization of the working principle is shown in Fig. 14(b). When the motor induces an abduction/adduction, the parallelograms deform in such a way that the rotation between the arm and hand brace coincides with the rotational center of the human wrist.

In a redesign of the Dampace elbow, Schorsch et al. [74] remarked that although the mechanism provides excellent fitting ability and is capable of reducing the biological joint torque, it does not allow for large forces to be transferred through the exoskeleton. This is a result of the passive joints that are present in the mechanism. In their work, they clarify the impact of this limitation by studying a lifting task of the elbow as shown in Fig. 15. Although the elbow joint is constructed with drums and a cable drive, in essence, it operates as a double parallelogram mechanism [89] and is represented as such in the drawing. The parallelogram mechanism is capable of balancing the torque created by the mass around the elbow joint, thus reducing muscle torque to zero.

However, by looking at the balancing of forces, it is clear that the parallelogram mechanism will leave a residual force, Fjoint (see Fig. 15), on the elbow proportional to the weight of the load [74]. Due to the reduction of muscle torque to zero, the residual joint force has changed magnitude and direction compared to unaugmented lifting [74,98]. Schorsch et al. [74] mention two potentially harmful consequences of this observation.

First, musculoskeletal injuries occur at a higher rate when the natural mechanics of a task are disrupted [99]. Second, in partial compensation of the muscle torques during movement, the creation of joint forces in the opposite direction will require different muscle activation patterns in order to produce a desired movement. This is unfortunate because nearly any change to the human musculoskeletal system or its pattern of coordination increases metabolic consumption of the human [21]. In order to address the issues mentioned here, Schorsch et al. [74] added a gravitation compensation system, specifically, an akr-type system as described by Tuijthof and Herder [100], onto the original elbow design. The akr-type system consists of two ideal linear springs, mounted onto the parallelogram mechanism as shown in Fig. 15. This mechanism can exert an upward force Fakr onto the human arm. By altering the parameters of the gravitation compensation system, e.g., the attachment locations and stiffnesses of the springs, the magnitude of the force can be regulated. By combining both systems, torque actuation can be provided in a way that compensates for misalignment of the joint axes and joint forces can be altered at will. The mechanical execution of this conceptual idea is shown in Fig. 15. Possible applications for this system include transferring forces to the ground through an exoskeleton in load bearing applications, reducing the load on damaged joints or increasing the load on bones to counter disuse atrophy [74].

Other researchers have also offered a solution for adding a force transfer possibility to the RRR mechanism. In the samsung-assist (or S-assist) exoskeleton [10], the RRR mechanism consists of three pulleys and two additional links as also seen in Ref. [89]. The same principle, shown in Fig. 14(a), is exploited to compensate for misalignment between the knee rotational joint and the exoskeleton. The proximal pulley, which is connected to the thigh frame, is remotely driven by a tendon-based actuator that is mounted onto the back of the wearer, along with the required batteries. This motion is transmitted through the other pulleys (until the distal one) to the shank by means of a cable connecting all pulleys. Because the pulleys connected to the thigh and shank frame can translate with respect to each other, movement of the knee flexion/extension axis with respect to the exoskeleton can be accommodated. In order to ensure comfortable interaction with human wearers while allowing the design to transmit forces through the structure, both aligning links have been connected with an additional structure: the body weight support structure. This ensures that the knee joint can support a portion of the wearer's weight, along with the mass of the exoskeleton by transmitting force to the ground. As a result, the wearer can lean on the S-assist without feeling the weight of the associated backpack [10]. The body weight support structure also prevents the mechanism from going into a singular configuration with all three pulleys aligned [10]. The design of Amigo et al. [58] can ensure the same functionality by actively actuating all three pulleys. However, these authors warned that the use of three actuators instead of one adds extra mass and inertia to the human limbs and can therefore not be used in assistive exoskeletons. Their mechanism is mainly aimed for use in rehabilitation exoskeletons, where the weight of the structure is supported from its base and increased energy consumption is not a problem [58]. An additional difficulty is that the motors should now be controlled to provide the required assistive torques, as well as the required movements, to ensure proper adaptation of the system to the movement of the knee axis of rotation. Their work only describes a theoretical model, so additional research is required to validate the performance of the system.

In Ref. [101], Saccares et al. present the iT-Knee, which, in contrast to the two devices described before, allows for full three-dimensional knee motion. Again, misalignment of the flexion/extension axis with the exoskeleton has been accounted for by the addition of two extra rotational DOFs in parallel with the flexion/extension joint of the exoskeleton. In the iT-knee, however, this is accomplished by two parallelogram structures in series with each other as previously also seen in the wrist of Beekhuis et al. [97]. This device presents extra joints that are responsible for allowing the knee joint motion in the frontal and transverse plane. Indeed, the knee internal/external rotation and the neutral abduction/adduction are allowed. Note that the mechanism allows for knee internal/external rotation only when the knee is flexed. However, this is in accordance with the anatomical function of the knee [102].

In the Robo-Mate project, a hip module consisting of an active flexion/extension joint and two nonactuated abduction/adduction axes in combination with a nonactuated ball joint was developed [103]. The ball joint links the thigh brace to the exoskeleton frame as seen in Fig. 16. The second abduction joint in combination with the ball joint connection to the brace is said to compensate for misalignment effects [104]. The three-dimensional ball joint operates as a hinge in a single plane. Thus, the ab-/adduction mechanism consists of three parallel hinges, compensating for misalignment as an RRR joint. This mechanism introduces a translational DOF into the system. Note that this translation, in combination with the one in the torso module (which is identical to the thigh module), also allows for a compensation of misalignment of the flexion/extension axis (R) because both RRR mechanisms fulfill the function of a P joint and are therefore operating as an RPP mechanism.

A similar mechanism is used in the passive back support exoskeleton by Näf et al. [90]. However, one difference to the RoboMate exoskeleton is that the order of joints is changed. The RRR mechanism is placed above the flexion–extension joint (See Fig. 9), potentially allowing for smaller link lengths because the hip joint axis of the exoskeleton is somewhat closer to that of the human. Additionally, this configuration allows the misalignment compensation mechanism to be used as a fitting mechanism. A prismatic joint is meant to prevent the thigh cuff from slipping up during lateral bending.

Other Configurations.

Note that the basic idea of the RRP, RPP, and RRR mechanisms is to allow the motions of the exoskeleton with respect to the human that are required to compensate for a misalignment of rotation axes. As such, translation of the passive DOFs unloads the human joints from undesired reaction forces due to misalignment, while the active rotational DOFs transfer the assistive torques onto the user. However, these methods only account for misalignment of a single rotation axis that does not change orientation.

Vitiello et al. [105] pointed out that the conventional kinematic redundancy methods do not take into account the laxity of the elbow articulation. The human elbow is a “loose hinge joint,” i.e., the rotation axis traces the surface of a conic double frustum over its motion range. To account for this movement, four passive DOFs were incorporated into the NEUROExos exoskeleton, providing the actuated flexion/extension DOF with two additional rotational and two translational DOFs [106]. The ROM of the passive mechanism allows the flexion/extension axis of the exoskeleton to trace a double frustum, whose outer dimensions satisfy both the intra- and intersubject variability of elbow axis laxity [105]. Note that the passive mechanism presented here is based on the natural kinematics elbow joint. Although it does not aim to fully mimic elbow kinematics, additional DOFs are selected with a particular joint motion in mind. Therefore, it is not universally applicable. Yet, due to the fact that its design was based on human joint kinematics, it can capture the full three-dimensional behavior of the elbow, unlike the designs in Refs. [74] and [97].

Another device that includes all three-dimensional movements of the elbow axis is the exoskeleton presented by Cai and Bidaud [107]. In essence, the kinematic design of the device is identical to the shoulder of the Dampace exoskeleton [39]: it includes three perpendicular rotational DOFs axes and allows for translation in three dimensions. However, in order to limit the mass and the cost of the device, translation in 3D space is allowed by the implementation of a parallel mechanism based on parallelogram structures as described in Ref. [108] and already proposed by Stienen et al. [89]. Despite the goal to limit the mass of the device, the prototype is rather bulky.

Celebi et al. [71] designed a knee exoskeleton, the ASSISTON-Knee, using a Schmidt-coupling between the actuator unit and the brace frames. The Schmidt-coupling allows for translations between the input and output shaft of the mechanism, in a plane perpendicular to the rotation axes. Therefore, it can accommodate the translation of the human knee joint axis during movement, thus compensating for any misalignment between the actuator and the knee. At the inner side of the leg, an RRR joint structure was mounted between the thigh and shank brace in order to stabilize the exoskeleton (see Fig. 17(a)).

Researchers in the same group were also responsible for the design of a novel three-RRP mechanism for the realization of a self-aligning knee joint [109]. The mechanism is shown in Fig. 17(b) and consists of three small actuators, whose torque is superimposed to actuate flexion/extension of the knee. All three actuators are mounted onto separate bodies (see Fig. 17(b) bodies S, T, and V) that share a common rotation axis (situated perpendicular to the base plate N, though point O). The three bodies are all connected to the output (E) through a revolute and prismatic joint (in P, Q, and R). The entire system has a total of 3DOF: Actuated rotation around the output axis of the system (perpendicular to output plate E, through point Z) and translation in the plane perpendicular to the rotation axis. Because the system employs three small actuators instead of one large actuator, it is highly back-drivable. This allows the joint axis to auto-align with the knee joint [71]. The same mechanism was later implemented in the ASSISTON-SE arm exoskeleton, where the translational DOFs allow for misalignment compensation of the flexion/extension axis, as well as shoulder depression/elevation and retraction/protraction [60].

Beil and Asfour [110] presented a nonanthropomorphic hip exoskeleton, accounting for all 3DOF of the hip, shown in Fig. 18, allowing flexion/extension, ab-/adduction, and internal/external rotation of the human hip joint. The former two are implemented with two hinges, indicated by L8 and L2, respectively. Eccentric clamps are used to make their position adjustable. The internal/external rotation is realized through a mechanism consisting of the three hinges L3, L5, and L7, and the two sliders L4 and L6. The combination of prismatic and revolute joints allows the flexion/extension joint to move on a circular trajectory around the anatomical hip center, thus enabling the internal/external rotation axis of the mechanism to align with the human one. Despite the need for manual alignment of flexion/extension and ab-/adduction joints, Beil and Asfour [110] declared that the ROM of the prototype is sufficient to allow its wearer to walk unhindered. They also remark that the results might be influenced by flexible deformation of the device structure.

In the load-bearing exoskeleton by Walsh et al. [28], a cam mechanism was implemented to account for misalignment of the hip ab-/adduction joint that is positioned next to the leg [111]. The mechanism is shown in Fig. 19. During leg ab-/adduction, the exoskeleton joint moves through a slot virtually lengthening/shortening the exoskeleton leg. This ensures that there is no displacement between the exoskeleton and the wearer along the limb.

Since the technical requirements for a metabolically beneficial, kinematically compatible hip exoskeleton are known (see Sec. 2.3), this section will assess the ability of state-of-the-art misalignment compensation techniques in meeting these requirements. In order to do this, all of the devices that were presented earlier are listed in Table 1 (see study limitations in Sec. 5.8).

We have grouped the investigated solutions depending on the misalignment compensation strategy that was used: manual alignment, addition of compliance in the brace/frame or joint, and kinematic redundancy following the RRP, RPP, RRR, or another principle. The different approaches were evaluated on their ability to compensate for misalignment; the volume and mass of the presented devices; the simplicity of the approach and its expandability to three dimensions; the ability of the joint mechanisms to transfer forces through the structure, e.g., transfer the weight of the exoskeleton to the ground; the relation between angle change of the exoskeleton and that of the human; and the donning procedure, i.e., required time and the possibility for donning the device autonomously.

In this section, the performance of the misalignment compensation techniques reviewed so far will be discussed extensively for each of the different evaluation criteria.

Misalignment Compensation Ability.

It is the opinion of the authors that the misalignment compensation ability of all the reviewed strategies is generally good. According to our evaluation, all the methods based on adding kinematic redundancy score very well on misalignment compensation capability. A slightly lower score was given to the RRR-type because the maximum misalignment that the devices in the review can compensate for is around 30–40 mm. Given that a misalignment of up to 10 cm is not out of the ordinary, this range is rather small. It is worth mentioning though that the joint structure was always used on the moving parts of the body, e.g., elbow, wrist or knee, and the mass and volume were therefore purposefully kept low. This unfortunately comes at the cost of reduced compensation abilities. Manual alignment and use of compliance score lower than the kinematic redundancy types because they cannot account for all the different contributors of misalignment and their performance on the others is highly dependent on the donning accuracy. Additionally, the performance of these methods is also dependent on the joint's comparability to a 1DOF hinge. Depending on the joint type, the compensation ability thus scores in between “ok” and “rather bad,” i.e., ±.

Volume, Mass, and Simplicity.

Exoskeleton mass and inertia have a significant effect on the metabolic cost of the user. This is exactly why minimizing them should be one of the main objectives in exoskeleton design [112]. Research for the knee joint has shown that the effects of added inertia on the kinematics of gait were significantly bigger than those of misalignment [52]. Therefore, it is not surprising that most researchers still depend on manual alignment, given that this technique scores very high on volume, mass, and simplicity, i.e., volume and mass of the devices are low and simplicity is high. Because of the simplicity of the joint structures—most devices use 1DOF hinge joints [7780,82]—it is much easier to keep the mass of the device low and keep it in close proximity of the wearer. The same is true for use of compliant elements. Note that this evaluation only takes the kinematic structure into account; mass, volume, and complexity of the actuators were left out of the equation.

The devices that use the RPP joint structure have the lowest score on volume and mass, although it should be mentioned that most of them are meant for rehabilitation. Because of this, limited to no effort has been spent in reducing the mass or the volume of the devices since most of the bulky and heavy components are located on a base frame. The device of Bartenbach et al. [11], which is stand-alone, is considerably less heavy and bulky than the other ones in the category. Use of the RRR and RRP joint structures results in exoskeletons that are reasonably sized and have a limited mass. The S-assist, for example, has been shown to fit underneath the clothing of the wearer [84]. Again, as mentioned earlier, this has a lot to do with the focus of the designer toward the application on more distal joints. The last category has a double score: ±. Due to the variability of the designs, it is difficult to provide a single score for the entire category. Some devices are very heavy and lead to bulky architecture [60,71,109], while some offer an elegant solution for the problem of misalignment [105,113]. The disadvantage of the latter is that they were specifically designed for the elbow and the design is therefore not directly applicable to other joints.

As for simplicity, the kinematic redundancy groups receive a mildly negative score, because they require the addition of extra components, complicating the joint design [43]. Additionally, the placement of passive elements results in an increased difficulty of statistical determination of the robot [42]. In Ref. [39], a statement is made that movement of passive elements is fully known after connection with the human. However, the soft connection, i.e., deformation of the flesh and so on, leaves several DOFs undetermined. Therefore, small deviations are possible and the system is not fully determined [32]. After tests with the Dampace exoskeleton, Stienen et al. [39] also recognized the presence of angular mismatch between the exoskeleton and the human limb due to the lack of interaction stiffness. The devices in the “other” category scored even lower, due to the large amount of elements that were used and the complexity of their design.

Expandability.

Most of the misalignment compensation strategies were demonstrated on 1DOF devices. For a joint like the hip, this needs to be expanded to a 3DOF design. Therefore, it is important to verify the ease with which this can be realized for each of the different approaches. For the manual alignment strategy this is hard, because some of the rotation axes of the human joints are located inside the human body, e.g., the hip rotation axis lies in the center of the thigh and goes through the torso in stance. Therefore, it is impossible to align the exoskeleton hinge, with the human DOF without the use of additional elements to change the position of the instantaneous rotation axis. When using compliance to account for misalignment, expansion to three DOFs is easily done by choosing a compliant element with a spherical architecture. Because the compliance of the brace/frame or joint structure can compensate for the inability to reach joint axes that are located inside the human body, no additional elements are required if the structure is compliant enough.

For kinematic redundancy systems, the RPP design is best equipped for easy expandability. The addition of two revolute joints, to allow rotations in three dimensions, and a single prismatic one to compensate for misalignment for all three rotations, is sufficient. Therefore, the three-dimensional system would consist of an RRRPPP joint. For the RRR configuration, some things need to be taken into consideration. In theory, the RRR joint allows rotations around one single axis and translation in the plane perpendicular to it. Similarly, to the RPP case, the design of a compensated three-dimensional joint would still require two additional R-DOFs to allow rotation and one translational DOF perpendicular to the existing translational plane. This can be ensured by adding another two R-DOFs, making a new RRR system with one of the rotation axes, thus allowing for a translational movement. However, the amount of compensation that the system can account for is limited due to the preference of keeping the mass and volume low. Therefore, it is not beneficial to use this limited range in order to compensate for misalignment in several directions as this would further limit the capabilities of the system. Thus, it is best to incorporate an RRR system for each of the rotational DOFs of the human joint. For the hip, this system would consist of a 9 R joint.

Again, for the other category, the performance of the devices is highly dependent on the individual design. The designs of Lenzi [113] and Vitiello et al. [105] were made exclusively for the elbow and are therefore fully equipped to deal with the three-dimensional nature of the elbow kinematics, assuring a compact exoskeleton device. The devices of Celebi et al. [71] and Ergin and Patoglu [60,109], on the other hand, are one-dimensional and should therefore be implemented three times for a three-dimensional joint, making the resulting mechanism extremely bulky and heavy. Therefore, the score on expandability can range anywhere from ++ to − −.

Force Transmission.

Transmitting force through the joint mechanism is easy in exoskeletons that employ manual alignment, because the joints are usually made with simple hinges. Compliant elements score well in this category, although they do require some design effort to make this possible. For example, the flexible frame that is used in the S-Assist [85] has been built in such a way that it can bend in order to allow tight fitting and small misalignment compensation. However, the designers have also ensured that it cannot buckle under the forces generated by the weight of the device, allowing this load to be transferred to the ground.

The kinematic redundancy approach is less suitable for transmitting forces through the joint mechanism. In most designs, the additional elements are fully passive and therefore move under influence of any force acting on the joint. This is exactly why they perform so efficiently in compensating the effects of misalignment: when a force acts on the mechanism due to a misalignment of the rotation axes, the passive DOFs move until the force is eliminated, thus unloading the human [39]. Nevertheless, one could implement additional elements into the joint design that do allow the transmission of forces. In the work of Schorsch et al. [74], a gravity compensation mechanism was added to the RRR elbow to compensate for the forces in the joint, generated by carrying a mass in the hands. Although this redesign allows such mechanisms to transmit forces, it usually comes at a cost of added mass and volume. For example: Amigo et al. [58] implemented the RRR system in a knee exoskeleton and provided an actuator for each rotational DOF, thus allowing the knee joint to transmit forces. However, they mentioned in their work that the increased mass and volume would make the device unfit for assistive applications. A good alternative solution was suggested by Bartenbach et al. [11]. By implementing the passive elements in between the exoskeleton frame and the braces, the weight of the device itself can still be transmitted to the ground through the rigid frame. On the other hand, because of the passive elements between brace and frame, weight of the subject cannot be routed through the exoskeleton without the addition of supplementary systems.

Angle Relation.

The goal of the manual alignment strategy is to ensure that the human and robot axes of rotation are aligned and act as a single rotation axis. Therefore, any change in angle of the exoskeleton joint corresponds to an identical change in angle of the human one. This greatly simplifies the design of the actuation system, as the state of the human can at any time be determined by measurement of the exoskeleton variables, e.g., angles, speeds. Thus, there is no need for sensors on the human limbs, which in turn reduces the mass and the complexity of the active system, i.e., the total exoskeleton with full actuator system included. However, because this feature is highly dependent on the obtained level of joint alignment, it is scored positive, yet not maximally for manual alignment. Compliant structures are scored mildly negative because their performance is dependent on the location of the structure with respect to the human joint. Additionally, due to their nature, movement of the exoskeleton induces deformation of the joint structure and as such a torque around the joint axis in the absence of any type of actuation [86,87]. Because the application of torques onto humans has already proven to have an influence on joint kinematics [56] and the presence of torques generally increases the deformation of soft tissues, the reliability compliant systems on angle relation is considered lower than that which employs manual alignment.

For the kinematically redundant systems, there is a high variability in performance for the angle relation between the device and the human. The RPP system scores very positive, due to the fact that there is only one rotational DOF present in the design. Because the other DOFs are translations, they cannot account for any rotation of the human joint; therefore, any rotation of the human limb must be allowed by a rotation of the exoskeleton joint. It deserves to be noted that a slight mismatch between angle changes can still be possible due to the deformation of soft tissues. However, this is true for each of the discussed approaches and therefore does not account for the changes in performance between the groups.

In the RRR system, the same can be achieved, although this requires a certain amount of design effort. Due to the nature of the system, a translation of both exoskeleton limbs with respect to each other is only possible by a rotation of the three hinges in the system. Therefore, without any additional systems, the rotation of the joint hinges cannot be directly coupled to a rotation of the human joint. However, when the system is equipped with an actuator, the actuation system can be executed in such a way that rotation of the actuator is independent of translation of the system, as mentioned in Ref. [53]. The state of the human can then be assessed based on angle measurements on the actuator. Because of this, the system is still positively evaluated when it comes to the angle relation, although slightly less positive than the RPP alternative.

In the RRP approach, there is an inherent angle deviation between the exoskeleton limb and the human when the human moves. The magnitude of this deviation is dependent on the position of the braces on the user and, as such, by the amount of misalignment. Although it is not possible to directly relate the angular deviation of the exoskeleton to that of its wearer, it could be possible to determine the relation between both by means of a calibration procedure. The downsides of this approach are that the procedures need to be repeated every single time that the device is donned and the sought relation can vary within a single operating period due to movement of the braces. For the devices in the other category, no single score can be given as the relation between angles is dependent on the design of the device. Therefore, for the group, the score can lie anywhere between ++ and − −.

Donning Procedure.

For a device to be easily used, the donning needs to be quick and easy. Preferably, the user should not require any assistance. For the first two categories, i.e., manual alignment and compliant elements, autonomous donning is as good as impossible. This is because donning requires an accurate localization of the rotation axes of the human joint, as well as an accurate placement of the device in order to best approximate the joint locations. Because localization of the joints is also dependent on several intrasubject variabilities, such as joint load, it is not possible to use predetermined values for the device parameters such as limb length and brace position. Therefore, the donning procedure is very lengthy.

The other mechanisms score very good on timing and difficulty of the donning procedure. Because misalignment is fully compensated for by the additional elements, it is no longer necessary to carefully position the joint mechanism onto the limb. Therefore, the user can just put the device on and is only required to make adaptations in order to improve his or her comfort. Because the mass and volume of most of the devices are reasonable, users are capable of donning them autonomously. In order to optimize autonomous donning, designers should make sure that all braces are closed the same way and on the same side of the limb, and the device can be easily put on like a pair of pants or a backpack. Only the RRR-based devices score one category lower because, due to focus on volume, the compensation range is rather limited. In order to ensure an optimal functioning of the mechanism, it is therefore preferable to mount it in close proximity to the human rotation axis.

Summary of Observations.

In the authors' opinion, manually aligned joints are difficult to align correctly with the respective joint of the subject. The misalignment compensation ability of this technique is highly dependent on the achieved alignment during donning, which requires the assistance of a skilled individual. Additionally, these techniques cannot account for all the different sources of misalignment and slip during operation has the potential to undo the alignment efforts.

As for kinematic redundancy methods, none of the devices in the other category perform well on meeting all the requirements. While some display excellent misalignment compensation abilities in a relatively light and compact package, they are designed for specific joints and would require a complete redesign for application at other joints. Additionally, the designs are usually complex and cannot be easily replaced in case of failure of a single component.

Due to the negative evaluation of the RRP architecture for expandability, it is not a suitable option. Expanding the design to a 3DOF system like the hip requires the inclusion of three separate RRP joint, with detrimental effects on mass and volume. Although mass and volume could be ameliorated by combining the three R-joints on the brace into a single spherical joint as seen in Ref. [103], the difficulties of linking the anatomical angle change to that of the exoskeleton exclude this option. This is mainly due to the increased need for additional sensors that are placed on the human limbs, which would increase the mass of the device as well as the complexity of donning it due to sensor placement.

At first glance, the RPP technique has terrible rating on volume and mass, which has a big influence on metabolic performance. However, it deserves to be noted that most RPP systems have been implemented in rehabilitation devices in which mass and volume are not an issue because heavy and bulky components are usually mounted on a fixed base.

The knee joint of the Bartenbach exoskeleton [11] shows that with some effort, mass and volume can be significantly reduced. In the case of volume, this can be achieved primarily by positioning sliding elements along the limbs and/or torso. For the RRR technique, devices were shown that fit underneath the clothes resulting in a good score on mass and volume. However, these good scores are generally at the cost of misalignment compensation ability as these designs require the use of small additional linkages between the revolute joints and, consequently, a reduced translational ROM. It follows that the expansion to a 3DOF system would best be done by implementing three separate RRR joints. Taking these arguments into consideration, the RPP architecture appears to be a valid approach for implementing kinematic compatibility in a metabolically beneficial exoskeleton on the condition that volume and mass of the joint structures are limited.

Study Limitations.

The authors want to mention that the data shown in Table 1 are based on an extensive literature research, authors' own expertise, and interpretation. Consequently, it should be regarded as an orientative rather than a definite guide. We have identified an existing gap in benchmarking that makes having scientifically strict conclusions elusive at the moment. Moreover, the advantages and disadvantages are closely dependent on the implementation; therefore, a compelling conclusion is still elusive and difficult to achieve.

Although, theoretically, kinematically compatible joint structures could be realized by mimicking human joint kinematics, in practice, this has been deemed infeasible. Because of large intrasubject differences, it is practically impossible to successfully mimic the joint kinematics of different subjects with a single mechanism. Even for a single subject, differences in muscle tension and joint loading would result in slightly different joint kinematics that cannot be captured by a single joint structure. Because of this, the common consensus is that the benefits of using complex joint models do not justify the added mass and complexity of the exoskeleton structures. Instead, researchers rely on simplified joint models that are a combination of hinges in various planes of motion.

Human joints are covered with soft tissue; therefore, the exact location of the joint axes is not clearly known. Estimation methods exist. However, in general, an error of several centimeters between the estimated rotation axis and the actual one is to be expected. In view of this fact, it is reasonable to assume that a misalignment between exoskeleton hinges and anatomical rotation axes will always be present. Unfortunately, misalignment of these rotation axes impedes the correct transfer of assistive torques onto the human and can potentially be harmful. Therefore, when using simplified joint models, it is necessary to include techniques that compensate for the effects of misalignment.

Misalignment compensation techniques usually fit within one of three categories: manual alignment, use of compliant elements, and the addition of kinematic redundancy. Manual alignment entails the adaptation of the exoskeleton during the donning process, in such a way that the rotation axes of the device align with the anatomical rotation axes of the human joint. However, due to joint coverage, alignment is never exact and the usefulness of the strategy is rather limited. Adding compliant elements alleviates the need for exact alignment. However, close alignment is still required because compensation abilities are limited. Kinematic redundancy methods perform better as they are capable of fully compensating for any misalignment between the rotation axes. However, this often comes at the cost of increased complexity, volume, and mass of the device.

In the kinematics redundancy-based architectures, the devices in the ‘other’ category provide adequate misalignment compensation, yet their complexity and the inability to easily expand the design to a three-dimensional joint make their use less attractive. Therefore, excluding the ‘other’ category leaves the RPP, RRP, and RRR kinematic compensation strategies. The RPP devices generally score bad on volume and mass, while RRP-based devices score bad on angle relation and expandability. RRR-based devices tend to score bad on simplicity, expandability, and ability to transmit forces but have reasonable scores in their ability to compensate for misalignment, volume, and mass.

All in all, no single method to compensate for misalignment fulfills all evaluated criteria without weaknesses in another. Therefore, despite significant progress in this field, designers of exoskeletons still have to choose, where they put their focus. In many cases, the focus might still be on low volume and mass, either with manual alignment or compliance either on brace or joint level. In other cases, the focus might be on the ability to compensate for misalignment and thereby increase comfort of the exoskeleton and reduce the donning and doffing time. In that case, exoskeleton designers might turn toward methods, where additional kinematic structures are introduced.

The work presented in this paper was supported by a grant from the Flemish Agency for Innovation by Science and Technology for the MIRAD project (MIRAD, IWT-SBO 120057), in conjunction with a funding from the European Commissions as part of the project SPEXOR under grant no. 6876623 and the Research Foundation - Flanders (FWO) under grant no. S000118N SBO Exo4Work project.

  • Agentschap Innoveren en Ondernemen (120057; Funder ID: 10.13039/100012331).

  • Fonds Wetenschappelijk Onderzoek (S000118N; Funder ID: 10.13039/501100003130).

  • Horizon 2020 (687662; Funder ID: 10.13039/501100007601).

Zoss, A. B. , Kazerooni, H. , and Chu, A. , 2006, “ Biomechanical Design of the Berkeley Lower Extremity Exoskeleton (bleex),” IEEE/ASME Trans. Mechatronics, 11(2), pp. 128–138. [CrossRef]
de Looze, M. P. , Bosch, T. , Krause, F. , Stadler, K. S. , and O'Sullivan, L. W. , 2016, “ Exoskeletons for Industrial Application and Their Potential Effects on Physical Work Load,” Ergonomics, 59(5), pp. 671–681. [CrossRef] [PubMed]
Näf, M. B. , De Rijcke, L. , Guerrero, C. R. , Millard, M. , Vanderborght, B. , and Lefeber, D. , 2017. “ Towards Low Back Support With a Passive Biomimetic Exo-Spine,” International Conference on Rehabilitation Robotics (ICORR), London, July 14–17, pp. 1165–1170.
Louie, D. R. , and Eng, J. J. , 2016, “ Powered Robotic Exoskeletons in Post-Stroke Rehabilitation of Gait: A Scoping Review,” J. Neuroeng. Rehabil., 13(1), p. 53. [CrossRef] [PubMed]
Kintsch, A. , and Depaula, R. , 2002, “ A Framework for the Adoption of Assistive Technology,” SWAAAC 2002: Supporting Learning Through Assistive Technology, Winter Park, CO, pp. 1–11.
Scherer, M. J. , 1996, “ Outcomes of Assistive Technology Use on Quality of Life,” Disability Rehabil., 18(9), pp. 439–448. [CrossRef]
Phillips, B. , and Zhao, H. , 1993, “ Predictors of Assistive Technology Abandonment,” Assistive Technol., 5(1), pp. 36–45. [CrossRef]
Pons, J. L. , 2008, Wearable Robots, Wiley, Ltd, Chichester, UK.
Cenciarini, M. , and Dollar, A. M. , 2011, “ Biomechanical Considerations in the Design of Lower Limb Exoskeletons,” IEEE International Conference on Rehabilitation Robotics, Zurich, Switzerland, June 29–July 1, pp. 10–14.
Choi, B. , Lee, Y. , Kim, J. , Lee, M. , Lee, J. , Roh, S. G. , Choi, H. , Kim, Y. J. , and Choi, J.-Y. , 2016, “ A Self-Aligning Knee Joint for Walking Assistance Devices,” Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), Orlando, FL, Aug. 16–20, pp. 2222–2227.
Bartenbach, V. , Wyss, D. , Seuret, D. , and Riener, R. , 2015, “ A Lower Limb Exoskeleton Research Platform to Investigate Human-Robot Interaction,” IEEE International Conference on Rehabilitation Robotics, Singapore, Aug. 11–14, pp. 600–605.
Kitatani, R. , Ohata, K. , Takahashi, H. , Shibuta, S. , Hashiguchi, Y. , and Yamakami, N. , 2014, “ Reduction in Energy Expenditure During Walking Using an Automated Stride Assistance Device in Healthy Young Adults,” Arch. Phys. Med. Rehabil., 95(11), pp. 2128–2133. [CrossRef] [PubMed]
Gams, A. , Petric, T. , Debevec, T. , and Babic, J. , 2013, “ Effects of Robotic Knee Exoskeleton on Human Energy Expenditure,” IEEE Trans. Biomed. Eng., 60(6), pp. 1636–1644. [CrossRef] [PubMed]
Farris, R. J. , Quintero, H. A. , and Goldfarb, M. , 2011, “ Preliminary Evaluation of a Powered Lower Limb Orthosis to Aid Walking in Paraplegic Individuals,” IEEE Trans. Neural Syst. Rehabil. Eng., 19(6), pp. 652–659. [CrossRef] [PubMed]
Ferris, D. P. , Sawicki, G. S. , and Daley, M. A. , 2007, “ A Physiologist's Perspective on Robotic Exoskeletons for Human Locomotion,” Int. J. HR: Humanoid Rob., 4(3), pp. 507–528. [CrossRef]
Yu, S. , Han, C. , and Cho, I. , 2014, “ Design Considerations of a Lower Limb Exoskeleton System to Assist Walking and Load-Carrying of Infantry Soldiers,” Appl. Bionics Mech., 11(3), pp. 119–134. [CrossRef]
Mooney, L. M. , and Herr, H. M. , 2016, “ Biomechanical Walking Mechanisms Underlying the Metabolic Reduction Caused by an Autonomous Exoskeleton,” J. Neuroeng. Rehabil., 13(4), pp. 1–12. [PubMed]
Sawicki, G. S. , and Ferris, D. P. , 2008, “ Mechanics and Energetics of Level Walking With Powered Ankle Exoskeletons,” J. Exp. Biol., 211(9), pp. 1402–1413. [CrossRef] [PubMed]
Sawicki, G. S. , and Ferris, D. P. , 2009, “ Mechanics and Energetics of Incline Walking With Robotic Ankle Exoskeletons,” J. Exp. Biol., 212(1), pp. 32–41. [CrossRef] [PubMed]
Malcolm, P. , Derave, W. , Galle, S. , and De Clercq, D. , 2013, “ A Simple Exoskeleton That Assists Plantarflexion Can Reduce the Metabolic Cost of Human Walking,” PLoS ONE, 8(2), p. e56137.
Collins, S. H. , Wiggin, M. B. , and Sawicki, G. S. , 2015, “ Reducing the Energy Cost of Human Walking Using an Unpowered Exoskeleton,” Nature, 522(7555), pp. 212–215. [CrossRef] [PubMed]
Mooney, L. M. , Rouse, E. J. , and Herr, H. M. , 2014, “ Autonomous Exoskeleton Reduces Metabolic Cost of Human Walking During load carriage,” J. Neuroeng. Rehabil., 11(1), p. 80. [CrossRef] [PubMed]
Mooney, L. M. , Rouse, E. J. , and Herr, H. M. , 2014, “ Autonomous Exoskeleton Reduces Metabolic Cost of Human Walking,” J. Neuroeng. Rehabil., 11(151), p. 5. [PubMed]
Seo, K. , Lee, J. , Lee, Y. , Ha, T. , and Shim, Y. , 2016, “ Fully Autonomous Hip Exoskeleton Saves Metabolic Cost of Walking,” IEEE International Conference on Robotics and Automation (ICRA), Stockholm, Sweden, May 16–21, pp. 4628–4635.
van Dijk, W. , and Van der Kooij, H. , 2014, “ XPED2: A Passive Exoskeleton With Artificial Tendons,” IEEE Rob. Autom. Mag., 21(4), pp. 56–61. [CrossRef]
Gregorczyk, K. N. , Hasselquist, L. , Schiffman, J. M. , Bensel, C. K. , Obusek, J. P. , and Gutekunst, D. J. , 2010, “ Effects of a Lower-Body Exoskeleton Device on Metabolic Cost and Gait Biomechanics During Load Carriage,” Ergonomics, 53(10), pp. 1263–1275. [CrossRef] [PubMed]
Asbeck, A. T. , De Rossi, S. M. M. , Holt, K. G. , and Walsh, C. J. , 2015, “ A Biologically Inspired Soft Exosuit for Walking Assistance,” Int. J. Rob. Res., 34(6), pp. 744–762. [CrossRef]
Walsh, C. J. , Endo, K. , and Herr, H. , 2007, “ A Quasi-Passive Leg Exoskeleton for Load-Carrying Augmentation,” Int. J. Humanoid Rob., 4(3), pp. 487–506. [CrossRef]
Galle, S. , 2015, “ Ankle-Foot Exoskeletons,” Ph.D. thesis, Universiteit Gent, Gent, Belgium.
Norman, D. A. , 1993, Things That Make us Smart: Defending Human Attributes in the Age of the Machine, Basic Books, New York.
Crowell, H. P. , 1995, “ Human Engineering Design Guidelines for a Powered, Full Body Exoskeleton,” Army Research Laboratory, Aberdeen Proving Ground, Aberdeen, MD, Report No. ARL-TN-60.
Lagoda, C. , Moreno, J. C. , and Pons, J. L. , 2012, “ Human-Robot Interfaces in Exoskeletons for Gait Training After Stroke: State of the Art and Challenges,” Appl. Bionics Mech., 9(2), pp. 193–203.
Bertomeu, J. M. B. , Lois, J. M. B. , Guillem, R. B. , Del Pozo, Á. P. , Lacuesta, J. , Mollà, C. G. , Luna, P. V. , and Pastor, J. P. , 2007, “ Development of a Hinge Compatible With the Kinematics of the Knee Joint,” Prosthet. Orthotics Int., 31(4), pp. 371–383.
Tucker, M. R. , Moser, A. , Lambercy, O. , Sulzer, J. , and Gassert, R. , 2013, “ Design of a Wearable Perturbator for Human Knee Impedance Estimation During Gait,” IEEE International Conference on Rehabilitation Robotics (ICORR), Seattle, WA, June 24–26, pp. 1–6.
Schiele, A. , and Van Der Helm, F. C. T. , 2006, “ Kinematic Design to Improve Ergonomics in Human Machine Interaction,” IEEE Trans. Neural Syst. Rehabil. Eng., 14(4), pp. 456–469. [PubMed]
Davis, R. B. , Õunpuu, S. , Tyburski, D. , and Gage, J. R. , 1991, “ A Gait Analysis Data Collection and Reduction Technique,” Hum. Mov. Sci., 10(5), pp. 575–587.
Seidel, G. K. , Marchinda, D. M. , Dijkers, M. , and Soutas-Little, R. W. , 1995, “ Hip Joint Center Location From Palpable Bony Landmarks-A Cadaver Study,” J. Biomech., 28(8), pp. 995–998. [PubMed]
Kirkwood, R. N. , Culham, E. G. , and Costigan, P. , 1999, “ Radiographic and Non-Invasive Determination of the Hip Joint Center Location: Effect on Hip Joint Moments,” Clin. Mech., 14(4), pp. 227–235. https://www.sciencedirect.com/science/article/pii/S0268003398000734
Stienen, A. H. A. , Hekman, E. E. G. , Prange, G. B. , Jannink, M. J. A. , Aalsma, A. M. M. , van der Helm, F. C. T. , and van der Kooij, H. , 2009, “ Dampace: Design of an Exoskeleton for Force-Coordination Training in Upper-Extremity Rehabilitation,” ASME J. Med. Devices, 3(3), p. 031003.
Cempini, M. , De Rossi, S. M. M. , Lenzi, T. , Vitiello, N. , and Carrozza, M. C. , 2013, “ Self-Alignment Mechanisms for Assistive Wearable Robots: A Kinetostatic Compatibility Method,” IEEE Trans. Rob., 29(1), pp. 236–250.
Pons, J. L. , 2010, “ Rehabilitation Exoskeletal Robotics,” IEEE Eng. Med. Biol. Mag., 29(3), pp. 57–63. [PubMed]
Nef, T. , Guidali, M. , and Riener, R. , 2009, “ ARMin III—Arm Therapy Exoskeleton With an Ergonomic Shoulder Actuation,” Appl. Bionics Mech., 6(2), pp. 127–142.
Gálvez-Zúñiga, M. A. , and Aceves-López, A. , 2016, “ A Review on Compliant Joint Mechanisms for Lower Limb Exoskeletons,” J. Rob., 2016, p. 5751391.
Wu, G. , Siegler, S. , Allard, P. , Kirtley, C. , Leardini, A. , Rosenbaum, D. , Whittle, M. , D'Lima, D. D. , Cristofolini, L. , Witte, H. , Schmid, O. , and Stokes, I. , 2002, “ ISB Recommendation on Definitions of Joint Coordinate System of Various Joints for the Reporting of Human Joint Motion—Part I: Ankle, Hip, and Spine,” J. Biomech., 35(4), pp. 543–548. [PubMed]
Kadaba, M. P. , Ramakrishnan, H. K. , and Wootten, M. E. , 1990, “ Measurement of Lower Extremity Kinematics During Level Walking,” J. Orthop. Res., 8(3), pp. 383–392. [PubMed]
Jarrassé, N. , and Morel, G. , 2012, “ Connecting a Human Limb to an Exoskeleton,” IEEE Trans. Rob., 28(3), pp. 697–709.
Schiele, A. , 2008, “ An Explicit Model to Predict and Interpret Constraint Force Creation in Phri With Exoskeletons,” IEEE International Conference on Robotics and Automation, Pasadena, CA, May 19–23, pp. 1324–1330.
Colombo, G. , Jorg, M. , Schreier, R. , and Dietz, V. , 2000, “ Treadmill Training of Paraplegic Patients Using a Robotic Orthosis,” J. Rehabil. Res. Dev., 37(6), pp. 693–700. https://pdfs.semanticscholar.org/0cb4/efa47bffffb6ffd44b6025f11184e919ed8d.pdf [PubMed]
Colombo, G. , Jorg, M. , and Dietz, V. , 2000, “ Driven Gait Orthosis to Do Locomotor Training of Paraplegic Patients,” 22nd Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Chicago, IL, July 23–28, pp. 3159–3163.
Accoto, D. , Sergi, F. , Tagliamonte, N. L. , Sudano, A. , and Guglielmelli, E. , 2014, “ A Nonanthropomorphic Wearable Robot,” IEEE Rob. Autom. Mag., 21(4), pp. 45–55.
Schiele, A. , 2008, “ Fundamentals of Ergonomic Exoskeleton Robots,” Ph.D. thesis, Delft, The Netherlands. https://repository.tudelft.nl/islandora/object/uuid%3A80d212ff-37d0-4ddd-aee1-820bb24cf35e
Zanotto, D. , Akiyama, Y. , Stegall, P. , and Agrawal, S. K. , 2015, “ Knee Joint Misalignment in Exoskeletons for the Lower Extremities: Effects on User's Gait,” IEEE Trans. Rob., 31(4), pp. 978–987.
Schiele, A. , and van der Helm, F. C. , 2009, “ Influence of Attachment Pressure and Kinematic Configuration on pHRI With Wearable Robots,” Appl. Bionics Mech., 6(2), pp. 157–173.
Schiele, A. , 2009, “ Ergonomics of Exoskeletons: Objective Performance Metrics,” Third Joint Eurohaptics Conference and Symposium on Haptic Interfaces for Virtual Environment and Teleoperator Systems, Salt Lake City, UT, Mar. 18–20, p. 6.
Hidler, J. M. , and Wall, A. E. , 2005, “ Alterations in Muscle Activation Patterns During Robotic-Assisted Walking,” Clin. Mech., 20(2), pp. 184–193. https://www.ncbi.nlm.nih.gov/pubmed/15621324
Hidler, J. , Wisman, W. , and Neckel, N. , 2008, “ Kinematic Trajectories While Walking Within the Lokomat Robotic Gait-Orthosis,” Clin. Mech., 23(10), pp. 1251–1259.
Nichols, D. , Neckel, N. D. , and Hidler, J. M. , 2006, “ Joint Moments Exhibited by Chronic Stroke Subjects While Walking With a Prescribed Physiological Gait Pattern,” J. Neurol. Phys. Ther., 30(4), pp. 220–221.
Amigo, L. E. , Casals, A. , and Amat, J. , 2011, “ Design of a 3-DoF Joint System With Dynamic Servo-Adaptation in Orthotic Applications,” IEEE International Conference on Robotics and Automation (ICRA), Shanghai, China, May 9–13, pp. 3700–3705.
Nef, T. , Mihelj, M. , Colombo, G. , and Riener, R. , 2006, “ ARMin-Robot for Rehabilitation of the Upper Extremities,” IEEE International Conference on Robotics and Automation, Orlando, FL, May 15–19, pp. 3152–3157.
Ergin, M. A. , and Patoglu, V. , 2012, “ ASSISTON-SE: A Self-Aligning Shoulder-Elbow Exoskeleton,” IEEE International Conference on Robotics and Automation, Saint Paul, MN, May 14–18, pp. 1579–1585.
Goodfellow, J. , and O'Connor, J. , 1978, “ The Mechanics of the Knee and Prosthesis Design,” J. Bone Jt. Surg. Br., 60-B(3), pp. 358–69.
Della Croce, U. , Leardini, A. , Chiari, L. , and Cappozzo, A. , 2005, “ Human Movement Analysis Using Stereophotogrammetry—Part 4: Assessment of Anatomical Landmark Misplacement and Its Effects on Joint Kinematics,” Gait Posture, 21(2), pp. 226–237. [PubMed]
Bell, A. L. , Pedersen, D. R. , and Brand, R. A. , 1990, “ A Comparison of the Accuracy of Several Hip Center Location Prediction Methods,” J. Biomech., 23(6), pp. 617–621. [PubMed]
Harman, E. , Han, K. H. , Frykman, P. , and Pandorf, C. , 2000, “ The Effects of Backpack Weight on the Biomechanics of Load Carriage,” United States Army Research Institute of Environmental Medicine, Natick, MA, Report No. USARIEM-T00-17.
Browning, R. C. , Modica, J. R. , Kram, R. , and Goswami, A. , 2007, “ The Effects of Adding Mass to the Legs on the Energetics and Biomechanics of Walking,” Med. Sci. Sports Exercise, 39(3), pp. 515–525.
Regufe, L. , Duarte, R. , Ramos, A. , Nadeau, J.-P. , Perry, N. , and Mesnard, M. , 2017, “ An Exhaustive Method for Researching Articular Orthosis Mechanisms at the Conceptual Design Stage,” Proc. CIRP, 60, pp. 482–487.
Riemer-Reiss, M. L. , and Wacker, R. R. , 2000, “ Factors Associated With Assistive Technology Discontinuance Among Individuals With Disabilities,” J. Rehabil., 66(3), p. 7.
Batavia, A. , and Hammer, G. , 1989, “ Consumer Criteria for Evaluating Assistive Devices: Implications for Technology Transfer,” 12th Annual Conference on Rehabilitation Technology, New Orleans, LA, pp. 194–195.
Swinnen, E. , Lafosse, C. , Van Nieuwenhoven, J. , Ilsbroukx, S. , Beckwee, D. , and Kerckhofs, E. , 2015, “ Neurological Patients and Their Lower Limb Orthotics: An Observational Pilot Study About Acceptance and Satisfaction,” Prosthet. Orthot. Int., 41(1), pp. 41–50.
Wessels, R. , Dijcks, B. , Soede, M. , Gelderblom, G. J. , and De Witte, L. , 2003, “ Non-Use of Provided Assistive Technology Devices, a Literature Overview,” Technol. Disability, 15(4), pp. 231–238.
Celebi, B. , Yalcin, M. , and Patoglu, V. , 2013, “ AssistOn-Knee: A Self-Aligning Knee Exoskeleton,” IEEE/RSJ International Conference on Intelligent Robots and Systems, Tokyo, Japan, Nov. 3–7, pp. 996–1002.
Neckel, N. , Wisman, W. , and Hidler, J. , 2006, “ Limb Alignment and Kinematics Inside a Lokomat Robotic Orthosis,” Annual International Conference of the IEEE Engineering in Medicine and Biology, New York, Aug. 30–Sept. 3, pp. 2698–2701.
Tao, G. D. , 2010, “ Mechanical Bracing Solutions to Decrease Tibial Slippage of Anklebot,” Ph.D., Massachusetts Institute of Technology, Cambridge, MA.
Schorsch, J. F. , Keemink, A. Q. L. , Stienen, A. H. A. , van der Helm, F. C. T. , and Abbink, D. A. , 2014, “ A Novel Self-Aligning Mechanism to Decouple Force and Torques for a Planar Exoskeleton Joint,” Mech. Sci., 5(2), pp. 29–35.
Mirbagheri, M. M. , Niu, X. , Kindig, M. , and Varoqui, D. , 2012, “ The Effects of Locomotor Training With a Robotic-Gait Orthosis (Lokomat) on Neuromuscular Properties in Persons With Chronic SCI,” Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), San Diego, CA, Aug. 28–Sept. 1, pp. 3854–3857.
Meuleman, J. , Van Asseldonk, E. H. F. , and Van Der Kooij, H. , 2013, “ Novel Actuation Design of a Gait Trainer With Shadow Leg Approach,” IEEE International Conference on Rehabilitation Robotics, Seattle, WA, June 24–26, pp. 1–8.
Jezernik, S. , Colombo, G. , Keller, T. , Frueh, H. , and Morari, M. , 2003, “ Robotic Orthosis Lokomat: A Rehabilitation and Research Tool,” Neuromodulation, 6(2), pp. 108–115. [PubMed]
Hocoma, 2017, “ Technical Data Lokomat® Pro,” Volketswil, Switzerland, accessed June 20, 2017, http://www.hocoma.com/fileadmin/user/Dokumente/ Lokomat/TECH_L6_140317_en.pdf
Cyberdyne Inc., 2017, “ What's HAL®?,” Tsukuba, Japan, accessed June 20, 2017, http://www.cyberdyne.jp/english/products/ LowerLimb_medical.html
Guizzo, E. , and Goldstein, H. , 2005, “ The Rise of the Body Bots,” IEEE Spectrum, 42(10), pp. 50–56.
Strickland, E. , 2012, “ Good-Bye, Wheelchair,” IEEE Spectrum, 49(1), pp. 30–32.
Esquenazi, A. , Talaty, M. , Packel, A. , and Saulino, M. , 2012, “ The ReWalk Powered Exoskeleton to Restore Ambulatory Function to Individuals With Thoracic-Level Motor-Complete Spinal Cord Injury,” Am. J. Phys. Med. Rehabil., 91(11), pp. 911–921. [PubMed]
Guerrero, C. R. , Grosu, V. , Grosu, S. , Leu, A. , Ristic-Durrant, D. , Vanderborght, B. , and Lefeber, D. , 2015, “ Torque Control of a Push-Pull Cable Driven Powered Orthosis for the Corbys Platform,” IEEE International Conference on Rehabilitation Robotics (ICORR), Singapore, Aug. 11–14, pp. 25–30.
Lee, Y. , Kim, Y.-J. , Lee, J. , Lee, M. , Choi, B. , Kim, J. , and Park, Y. J. , 2017, “ Biomechanical Design of a Novel Flexible Exoskeleton for Lower Extremities,” IEEE/ASME Trans. Mechatronics, 22(5), pp. 2058–2069.
Lee, Y. , Choi, B. , Lee, J. , Lee, M. , Roh, S.-G. , Kim, J. , Choi, H. , and Kim, Y.-J. , 2016, “ Flexible Sliding Frame for Gait Enhancing Mechatronic System (GEMS),” IEEE 38th Annual International Conference of the Engineering in Medicine and Biology Society (EMBC), Orlando, FL, Aug. 16–20, pp. 598–602.
Kleinjan, J. G. , Dunning, A. G. , and Herder, J. L. , 2014, “ Design of a Compact Actuated Compliant Elbow Joint,” Int. J. Struct. Stab. Dyn., 14(8), p. 1440030.
Machekposhti, F. D. , Tolou, N. , and Herder, J. L. , 2015, “ A Review on Compliant Joints and Rigid-Body Constant Velocity Universal Joints Toward the Design of Compliant Homokinetic Couplings,” ASME J. Mech. Des., 137(3), p. 032301.
Schiele, A. , and Visentin, G. , 2003, “ The ESA Human Arm Exoskeleton for Space Robotics Telepresence,” Seventh International Symposium on Artificial Intelligence, Robotics and Automation in Space, Nara, Japan, May 19–23, pp. 19–23. https://pdfs.semanticscholar.org/86fd/abd1517902b73dc9cd85c172726baafd012f.pdf
Stienen, A. H. A. , Hekman, E. E. G. , van der Helm, F. C. T. , and van der Kooij, H. , 2009, “ Self-Aligning Exoskeleton Axes Through Decoupling of Joint Rotations and Translations,” IEEE Trans. Rob., 25(3), pp. 628–633.
Näf, M. B. , Koopman, A. S. , Baltrusch, S. , Rodriguez-Guerrero, C. , Vanderborght, B. , and Lefeber, D. , 2018, “ Passive Back Support Exoskeleton Improves Range of Motion Using Flexible Beams,” Front. Rob. AI, 5, p. 72.
d'Elia, N. , Vanetti, F. , Cempini, M. , Pasquini, G. , Parri, A. , Rabuffetti, M. , Ferrarin, M. , Lova, R. M. , and Vitiello, N. , 2017, “ Physical Human-Robot Interaction of an Active Pelvis Orthosis: Toward Ergonomic Assessment of Wearable Robots,” J. Neuroeng. Rehabil., 14(1), p. 29. [PubMed]
Cai, V. A. D. , Bidaud, P. , Hayward, V. , Gosselin, F. , and Desailly, E. , 2011, “ Self-Adjusting, Isostatic Exoskeleton for the Human Knee Joint,” Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), Boston, MA, Aug. 30–Sept. 3, pp. 612–618.
Cai, V. A. D. , Bidaud, P. , and Nguyen, V. T. , 2016, “ Control of a Self-Adjusting Lower Limb Exoskeleton for Knee Assistance,” Robot Design, Dynamics And Control (ROMANSY 21), Udine, Italy, June 20–23, pp. 385–392.
Mihelj, M. , Nef, T. , and Riener, R. , 2007, “ ARMin II—7 DoF Rehabilitation Robot: Mechanics and Kinematics,” IEEE International Conference on Robotics and Automation, Roma, Italy, Apr. 10–14, pp. 4120–4125.
Park, H. S. , Ren, Y. , and Zhang, L. Q. , 2008, “ IntelliArm: An Exoskeleton for Diagnosis and Treatment of Patients With Neurological Impairments,” Second Biennial IEEE/RAS-EMBS International Conference on Biomedical Robotics and Biomechatronics (BioRob), Scottsdale, AZ, Oct. 19–22, pp. 109–114.
Otten, A. , Voort, C. , Stienen, A. , Aarts, R. , van Asseldonk, E. , and van der Kooij, H. , 2015, “ LIMPACT:A Hydraulically Powered Self-Aligning Upper Limb Exoskeleton,” IEEE/ASME Trans. Mechatronics, 20(5), pp. 2285–2298.
Beekhuis, J. H. , Westerveld, A. J. , Van Der Kooij, H. , and Stienen, A. H. A. , 2013, “ Design of a Self-Aligning 3-DOF Actuated Exoskeleton for Diagnosis and Training of Wrist and Forearm After Stroke,” IEEE International Conference on Rehabilitation Robotics (ICORR), Seattle, WA, June 24–26, pp. 1–5.
Stormont, T. J. , An, K. N. , Morrey, B. F. , and Chao, E. Y. , 1985, “ Elbow Joint Contact Study: Comparison of Techniques,” J. Biomech., 18(5), pp. 329–336. [PubMed]
Kumar, S. , 2001, “ Theories of Musculoskeletal Injury Causation,” Ergonomics, 44(1), pp. 17–47. [PubMed]
Tuijthof, G. J. M. , and Herder, J. L. , 2000, “ Design, Actuation and Control of an Anthropomorphic Robot Arm,” Mech. Mach. Theory, 35(7), pp. 945–962.
Saccares, L. , Sarakoglou, I. , and Tsagarakis, N. G. , 2016, “ iT-Knee: An Exoskeleton With Ideal Torque Transmission Interface for Ergonomic Power Augmentation,” IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Daejeon, South Korea, Oct. 9–14, pp. 780–786.
Asano, Y. , Mizoguchi, H. , Kozuki, T. , Motegi, Y. , Osada, M. , Urata, J. , Nakanishi, Y. , Okada, K. , and Inaba, M. , 2012, “ Lower Thigh Design of Detailed Musculoskeletal Humanoid 'Kenshiro,” IEEE International Conference on Intelligent Robots and Systems, Vilamoura, Portugal, Oct. 7–12, pp. 4367–4372.
Toxiri, S. , Ortiz, J. , Masood, J. , Fernández, J. , Mateos, L. A. , and Caldwell, D. G. , 2016, “ A Powered Low-Back Exoskeleton for Industrial Handling: Considerations on Controls,” Wearable Robotics: Challenges and Trends, Vol. 16, J. González-Vargas, J. Ibáñez, J. Contreras-Vidal, H. van der Kooij, and J. Pons, eds., Springer, Cham, Switzerland, pp. 287–291.
Toxiri, S. , Ortiz, J. , Masood, J. , Fernandez, J. , Mateos, L. A. , and Caldwell, D. G. , 2015, “ A Wearable Device for Reducing Spinal Loads During Lifting Tasks: Biomechanics and Design Concepts,” IEEE International Conference on Robotics and Biomimetics (ROBIO), Zhuhai, China, Dec. 6–9, pp. 2295–2300.
Vitiello, N. , Lenzi, T. , Roccella, S. , De Rossi, S. M. M. , Cattin, E. , Giovacchini, F. , Vecchi, F. , and Carrozza, M. C. , 2013, “ NEUROExos: A Powered Elbow Exoskeleton for Physical Rehabilitation,” IEEE Trans. Rob., 29(1), pp. 220–235.
Lenzi, T. , Vitiello, N. , De Rossi, S. M. M. , Roccella, S. , Vecchi, F. , and Carrozza, M. C. , 2011, “ NEUROExos: A Variable Impedance Powered Elbow Exoskeleton,” IEEE International Conference on Robotics and Automation, Shanghai, China, May 9–13, pp. 1419–1426.
Cai, V. A. D. , and Bidaud, P. , 2014, “ Self-Adjusting Isostatic Exoskeleton for the Elbow Joint: Mechanical Design Parallel Manipulators,” Advances on Theory and Practice of Robots and Manipulators: Proceedings of ROMANSY 2014 XX CISM-IFToMM Symposium on Theory and Practice of Robots and Manipulators, Springer, Cham, Switzerland, pp. 19–26.
Pierrot, F. , Reynaud, C. , and Fournier, A. , 1990, “ Delta: A Simple and Efficient Parallel Robot,” Robotica, 8(2), pp. 105–109.
Ergin, M. A. , and Patoglu, V. , 2011, “ A Self-Adjusting Knee Exoskeleton for Robot-Assisted Treatment of Knee Injuries,” IEEE International Conference on Intelligent Robots and Systems, San Francisco, CA, Sept. 25–30, pp. 4917–4922.
Beil, J. , and Asfour, T. , 2016, “ New Mechanism for a 3 DOF Exoskeleton Hip Joint With Five Revolute and Two Prismatic Joints,” IEEE RAS and EMBS International Conference on Biomedical Robotics and Biomechatronics, Singapore, June 26–29, pp. 787–792.
Valiente, A. , Herr, H. , and Blanco, E. , 2003, “ Design of a Quasi-Passive Parallel Leg Exoskeleton to Augment Load Carrying for Walking by,” Ph.D. thesis, MIT, Cambridge, MA. https://dspace.mit.edu/handle/1721.1/34412
Shamaei, K. , Cenciarini, M. , Adams, A. A. , Gregorczyk, K. N. , Schiffman, J. M. , and Dollar, A. M. , 2015, “ Biomechanical Effects of Stiffness in Parallel With the Knee Joint During Walking,” IEEE Trans. Biomed. Eng., 62(10), pp. 2389–2401. [PubMed]
Lenzi, T. , Zanotto, D. , Stegall, P. , Carrozza, M. C. , and Agrawal, S. K. , 2012, “ Reducing Muscle Effort in Walking Through Powered Exoskeletons,” Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), San Diego, CA, Aug. 28–Sept. 1, pp. 3926–3929.
Copyright © 2018 by ASME
View article in PDF format.

References

Zoss, A. B. , Kazerooni, H. , and Chu, A. , 2006, “ Biomechanical Design of the Berkeley Lower Extremity Exoskeleton (bleex),” IEEE/ASME Trans. Mechatronics, 11(2), pp. 128–138. [CrossRef]
de Looze, M. P. , Bosch, T. , Krause, F. , Stadler, K. S. , and O'Sullivan, L. W. , 2016, “ Exoskeletons for Industrial Application and Their Potential Effects on Physical Work Load,” Ergonomics, 59(5), pp. 671–681. [CrossRef] [PubMed]
Näf, M. B. , De Rijcke, L. , Guerrero, C. R. , Millard, M. , Vanderborght, B. , and Lefeber, D. , 2017. “ Towards Low Back Support With a Passive Biomimetic Exo-Spine,” International Conference on Rehabilitation Robotics (ICORR), London, July 14–17, pp. 1165–1170.
Louie, D. R. , and Eng, J. J. , 2016, “ Powered Robotic Exoskeletons in Post-Stroke Rehabilitation of Gait: A Scoping Review,” J. Neuroeng. Rehabil., 13(1), p. 53. [CrossRef] [PubMed]
Kintsch, A. , and Depaula, R. , 2002, “ A Framework for the Adoption of Assistive Technology,” SWAAAC 2002: Supporting Learning Through Assistive Technology, Winter Park, CO, pp. 1–11.
Scherer, M. J. , 1996, “ Outcomes of Assistive Technology Use on Quality of Life,” Disability Rehabil., 18(9), pp. 439–448. [CrossRef]
Phillips, B. , and Zhao, H. , 1993, “ Predictors of Assistive Technology Abandonment,” Assistive Technol., 5(1), pp. 36–45. [CrossRef]
Pons, J. L. , 2008, Wearable Robots, Wiley, Ltd, Chichester, UK.
Cenciarini, M. , and Dollar, A. M. , 2011, “ Biomechanical Considerations in the Design of Lower Limb Exoskeletons,” IEEE International Conference on Rehabilitation Robotics, Zurich, Switzerland, June 29–July 1, pp. 10–14.
Choi, B. , Lee, Y. , Kim, J. , Lee, M. , Lee, J. , Roh, S. G. , Choi, H. , Kim, Y. J. , and Choi, J.-Y. , 2016, “ A Self-Aligning Knee Joint for Walking Assistance Devices,” Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), Orlando, FL, Aug. 16–20, pp. 2222–2227.
Bartenbach, V. , Wyss, D. , Seuret, D. , and Riener, R. , 2015, “ A Lower Limb Exoskeleton Research Platform to Investigate Human-Robot Interaction,” IEEE International Conference on Rehabilitation Robotics, Singapore, Aug. 11–14, pp. 600–605.
Kitatani, R. , Ohata, K. , Takahashi, H. , Shibuta, S. , Hashiguchi, Y. , and Yamakami, N. , 2014, “ Reduction in Energy Expenditure During Walking Using an Automated Stride Assistance Device in Healthy Young Adults,” Arch. Phys. Med. Rehabil., 95(11), pp. 2128–2133. [CrossRef] [PubMed]
Gams, A. , Petric, T. , Debevec, T. , and Babic, J. , 2013, “ Effects of Robotic Knee Exoskeleton on Human Energy Expenditure,” IEEE Trans. Biomed. Eng., 60(6), pp. 1636–1644. [CrossRef] [PubMed]
Farris, R. J. , Quintero, H. A. , and Goldfarb, M. , 2011, “ Preliminary Evaluation of a Powered Lower Limb Orthosis to Aid Walking in Paraplegic Individuals,” IEEE Trans. Neural Syst. Rehabil. Eng., 19(6), pp. 652–659. [CrossRef] [PubMed]
Ferris, D. P. , Sawicki, G. S. , and Daley, M. A. , 2007, “ A Physiologist's Perspective on Robotic Exoskeletons for Human Locomotion,” Int. J. HR: Humanoid Rob., 4(3), pp. 507–528. [CrossRef]
Yu, S. , Han, C. , and Cho, I. , 2014, “ Design Considerations of a Lower Limb Exoskeleton System to Assist Walking and Load-Carrying of Infantry Soldiers,” Appl. Bionics Mech., 11(3), pp. 119–134. [CrossRef]
Mooney, L. M. , and Herr, H. M. , 2016, “ Biomechanical Walking Mechanisms Underlying the Metabolic Reduction Caused by an Autonomous Exoskeleton,” J. Neuroeng. Rehabil., 13(4), pp. 1–12. [PubMed]
Sawicki, G. S. , and Ferris, D. P. , 2008, “ Mechanics and Energetics of Level Walking With Powered Ankle Exoskeletons,” J. Exp. Biol., 211(9), pp. 1402–1413. [CrossRef] [PubMed]
Sawicki, G. S. , and Ferris, D. P. , 2009, “ Mechanics and Energetics of Incline Walking With Robotic Ankle Exoskeletons,” J. Exp. Biol., 212(1), pp. 32–41. [CrossRef] [PubMed]
Malcolm, P. , Derave, W. , Galle, S. , and De Clercq, D. , 2013, “ A Simple Exoskeleton That Assists Plantarflexion Can Reduce the Metabolic Cost of Human Walking,” PLoS ONE, 8(2), p. e56137.
Collins, S. H. , Wiggin, M. B. , and Sawicki, G. S. , 2015, “ Reducing the Energy Cost of Human Walking Using an Unpowered Exoskeleton,” Nature, 522(7555), pp. 212–215. [CrossRef] [PubMed]
Mooney, L. M. , Rouse, E. J. , and Herr, H. M. , 2014, “ Autonomous Exoskeleton Reduces Metabolic Cost of Human Walking During load carriage,” J. Neuroeng. Rehabil., 11(1), p. 80. [CrossRef] [PubMed]
Mooney, L. M. , Rouse, E. J. , and Herr, H. M. , 2014, “ Autonomous Exoskeleton Reduces Metabolic Cost of Human Walking,” J. Neuroeng. Rehabil., 11(151), p. 5. [PubMed]
Seo, K. , Lee, J. , Lee, Y. , Ha, T. , and Shim, Y. , 2016, “ Fully Autonomous Hip Exoskeleton Saves Metabolic Cost of Walking,” IEEE International Conference on Robotics and Automation (ICRA), Stockholm, Sweden, May 16–21, pp. 4628–4635.
van Dijk, W. , and Van der Kooij, H. , 2014, “ XPED2: A Passive Exoskeleton With Artificial Tendons,” IEEE Rob. Autom. Mag., 21(4), pp. 56–61. [CrossRef]
Gregorczyk, K. N. , Hasselquist, L. , Schiffman, J. M. , Bensel, C. K. , Obusek, J. P. , and Gutekunst, D. J. , 2010, “ Effects of a Lower-Body Exoskeleton Device on Metabolic Cost and Gait Biomechanics During Load Carriage,” Ergonomics, 53(10), pp. 1263–1275. [CrossRef] [PubMed]
Asbeck, A. T. , De Rossi, S. M. M. , Holt, K. G. , and Walsh, C. J. , 2015, “ A Biologically Inspired Soft Exosuit for Walking Assistance,” Int. J. Rob. Res., 34(6), pp. 744–762. [CrossRef]
Walsh, C. J. , Endo, K. , and Herr, H. , 2007, “ A Quasi-Passive Leg Exoskeleton for Load-Carrying Augmentation,” Int. J. Humanoid Rob., 4(3), pp. 487–506. [CrossRef]
Galle, S. , 2015, “ Ankle-Foot Exoskeletons,” Ph.D. thesis, Universiteit Gent, Gent, Belgium.
Norman, D. A. , 1993, Things That Make us Smart: Defending Human Attributes in the Age of the Machine, Basic Books, New York.
Crowell, H. P. , 1995, “ Human Engineering Design Guidelines for a Powered, Full Body Exoskeleton,” Army Research Laboratory, Aberdeen Proving Ground, Aberdeen, MD, Report No. ARL-TN-60.
Lagoda, C. , Moreno, J. C. , and Pons, J. L. , 2012, “ Human-Robot Interfaces in Exoskeletons for Gait Training After Stroke: State of the Art and Challenges,” Appl. Bionics Mech., 9(2), pp. 193–203.
Bertomeu, J. M. B. , Lois, J. M. B. , Guillem, R. B. , Del Pozo, Á. P. , Lacuesta, J. , Mollà, C. G. , Luna, P. V. , and Pastor, J. P. , 2007, “ Development of a Hinge Compatible With the Kinematics of the Knee Joint,” Prosthet. Orthotics Int., 31(4), pp. 371–383.
Tucker, M. R. , Moser, A. , Lambercy, O. , Sulzer, J. , and Gassert, R. , 2013, “ Design of a Wearable Perturbator for Human Knee Impedance Estimation During Gait,” IEEE International Conference on Rehabilitation Robotics (ICORR), Seattle, WA, June 24–26, pp. 1–6.
Schiele, A. , and Van Der Helm, F. C. T. , 2006, “ Kinematic Design to Improve Ergonomics in Human Machine Interaction,” IEEE Trans. Neural Syst. Rehabil. Eng., 14(4), pp. 456–469. [PubMed]
Davis, R. B. , Õunpuu, S. , Tyburski, D. , and Gage, J. R. , 1991, “ A Gait Analysis Data Collection and Reduction Technique,” Hum. Mov. Sci., 10(5), pp. 575–587.
Seidel, G. K. , Marchinda, D. M. , Dijkers, M. , and Soutas-Little, R. W. , 1995, “ Hip Joint Center Location From Palpable Bony Landmarks-A Cadaver Study,” J. Biomech., 28(8), pp. 995–998. [PubMed]
Kirkwood, R. N. , Culham, E. G. , and Costigan, P. , 1999, “ Radiographic and Non-Invasive Determination of the Hip Joint Center Location: Effect on Hip Joint Moments,” Clin. Mech., 14(4), pp. 227–235. https://www.sciencedirect.com/science/article/pii/S0268003398000734
Stienen, A. H. A. , Hekman, E. E. G. , Prange, G. B. , Jannink, M. J. A. , Aalsma, A. M. M. , van der Helm, F. C. T. , and van der Kooij, H. , 2009, “ Dampace: Design of an Exoskeleton for Force-Coordination Training in Upper-Extremity Rehabilitation,” ASME J. Med. Devices, 3(3), p. 031003.
Cempini, M. , De Rossi, S. M. M. , Lenzi, T. , Vitiello, N. , and Carrozza, M. C. , 2013, “ Self-Alignment Mechanisms for Assistive Wearable Robots: A Kinetostatic Compatibility Method,” IEEE Trans. Rob., 29(1), pp. 236–250.
Pons, J. L. , 2010, “ Rehabilitation Exoskeletal Robotics,” IEEE Eng. Med. Biol. Mag., 29(3), pp. 57–63. [PubMed]
Nef, T. , Guidali, M. , and Riener, R. , 2009, “ ARMin III—Arm Therapy Exoskeleton With an Ergonomic Shoulder Actuation,” Appl. Bionics Mech., 6(2), pp. 127–142.
Gálvez-Zúñiga, M. A. , and Aceves-López, A. , 2016, “ A Review on Compliant Joint Mechanisms for Lower Limb Exoskeletons,” J. Rob., 2016, p. 5751391.
Wu, G. , Siegler, S. , Allard, P. , Kirtley, C. , Leardini, A. , Rosenbaum, D. , Whittle, M. , D'Lima, D. D. , Cristofolini, L. , Witte, H. , Schmid, O. , and Stokes, I. , 2002, “ ISB Recommendation on Definitions of Joint Coordinate System of Various Joints for the Reporting of Human Joint Motion—Part I: Ankle, Hip, and Spine,” J. Biomech., 35(4), pp. 543–548. [PubMed]
Kadaba, M. P. , Ramakrishnan, H. K. , and Wootten, M. E. , 1990, “ Measurement of Lower Extremity Kinematics During Level Walking,” J. Orthop. Res., 8(3), pp. 383–392. [PubMed]
Jarrassé, N. , and Morel, G. , 2012, “ Connecting a Human Limb to an Exoskeleton,” IEEE Trans. Rob., 28(3), pp. 697–709.
Schiele, A. , 2008, “ An Explicit Model to Predict and Interpret Constraint Force Creation in Phri With Exoskeletons,” IEEE International Conference on Robotics and Automation, Pasadena, CA, May 19–23, pp. 1324–1330.
Colombo, G. , Jorg, M. , Schreier, R. , and Dietz, V. , 2000, “ Treadmill Training of Paraplegic Patients Using a Robotic Orthosis,” J. Rehabil. Res. Dev., 37(6), pp. 693–700. https://pdfs.semanticscholar.org/0cb4/efa47bffffb6ffd44b6025f11184e919ed8d.pdf [PubMed]
Colombo, G. , Jorg, M. , and Dietz, V. , 2000, “ Driven Gait Orthosis to Do Locomotor Training of Paraplegic Patients,” 22nd Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Chicago, IL, July 23–28, pp. 3159–3163.
Accoto, D. , Sergi, F. , Tagliamonte, N. L. , Sudano, A. , and Guglielmelli, E. , 2014, “ A Nonanthropomorphic Wearable Robot,” IEEE Rob. Autom. Mag., 21(4), pp. 45–55.
Schiele, A. , 2008, “ Fundamentals of Ergonomic Exoskeleton Robots,” Ph.D. thesis, Delft, The Netherlands. https://repository.tudelft.nl/islandora/object/uuid%3A80d212ff-37d0-4ddd-aee1-820bb24cf35e
Zanotto, D. , Akiyama, Y. , Stegall, P. , and Agrawal, S. K. , 2015, “ Knee Joint Misalignment in Exoskeletons for the Lower Extremities: Effects on User's Gait,” IEEE Trans. Rob., 31(4), pp. 978–987.
Schiele, A. , and van der Helm, F. C. , 2009, “ Influence of Attachment Pressure and Kinematic Configuration on pHRI With Wearable Robots,” Appl. Bionics Mech., 6(2), pp. 157–173.
Schiele, A. , 2009, “ Ergonomics of Exoskeletons: Objective Performance Metrics,” Third Joint Eurohaptics Conference and Symposium on Haptic Interfaces for Virtual Environment and Teleoperator Systems, Salt Lake City, UT, Mar. 18–20, p. 6.
Hidler, J. M. , and Wall, A. E. , 2005, “ Alterations in Muscle Activation Patterns During Robotic-Assisted Walking,” Clin. Mech., 20(2), pp. 184–193. https://www.ncbi.nlm.nih.gov/pubmed/15621324
Hidler, J. , Wisman, W. , and Neckel, N. , 2008, “ Kinematic Trajectories While Walking Within the Lokomat Robotic Gait-Orthosis,” Clin. Mech., 23(10), pp. 1251–1259.
Nichols, D. , Neckel, N. D. , and Hidler, J. M. , 2006, “ Joint Moments Exhibited by Chronic Stroke Subjects While Walking With a Prescribed Physiological Gait Pattern,” J. Neurol. Phys. Ther., 30(4), pp. 220–221.
Amigo, L. E. , Casals, A. , and Amat, J. , 2011, “ Design of a 3-DoF Joint System With Dynamic Servo-Adaptation in Orthotic Applications,” IEEE International Conference on Robotics and Automation (ICRA), Shanghai, China, May 9–13, pp. 3700–3705.
Nef, T. , Mihelj, M. , Colombo, G. , and Riener, R. , 2006, “ ARMin-Robot for Rehabilitation of the Upper Extremities,” IEEE International Conference on Robotics and Automation, Orlando, FL, May 15–19, pp. 3152–3157.
Ergin, M. A. , and Patoglu, V. , 2012, “ ASSISTON-SE: A Self-Aligning Shoulder-Elbow Exoskeleton,” IEEE International Conference on Robotics and Automation, Saint Paul, MN, May 14–18, pp. 1579–1585.
Goodfellow, J. , and O'Connor, J. , 1978, “ The Mechanics of the Knee and Prosthesis Design,” J. Bone Jt. Surg. Br., 60-B(3), pp. 358–69.
Della Croce, U. , Leardini, A. , Chiari, L. , and Cappozzo, A. , 2005, “ Human Movement Analysis Using Stereophotogrammetry—Part 4: Assessment of Anatomical Landmark Misplacement and Its Effects on Joint Kinematics,” Gait Posture, 21(2), pp. 226–237. [PubMed]
Bell, A. L. , Pedersen, D. R. , and Brand, R. A. , 1990, “ A Comparison of the Accuracy of Several Hip Center Location Prediction Methods,” J. Biomech., 23(6), pp. 617–621. [PubMed]
Harman, E. , Han, K. H. , Frykman, P. , and Pandorf, C. , 2000, “ The Effects of Backpack Weight on the Biomechanics of Load Carriage,” United States Army Research Institute of Environmental Medicine, Natick, MA, Report No. USARIEM-T00-17.
Browning, R. C. , Modica, J. R. , Kram, R. , and Goswami, A. , 2007, “ The Effects of Adding Mass to the Legs on the Energetics and Biomechanics of Walking,” Med. Sci. Sports Exercise, 39(3), pp. 515–525.
Regufe, L. , Duarte, R. , Ramos, A. , Nadeau, J.-P. , Perry, N. , and Mesnard, M. , 2017, “ An Exhaustive Method for Researching Articular Orthosis Mechanisms at the Conceptual Design Stage,” Proc. CIRP, 60, pp. 482–487.
Riemer-Reiss, M. L. , and Wacker, R. R. , 2000, “ Factors Associated With Assistive Technology Discontinuance Among Individuals With Disabilities,” J. Rehabil., 66(3), p. 7.
Batavia, A. , and Hammer, G. , 1989, “ Consumer Criteria for Evaluating Assistive Devices: Implications for Technology Transfer,” 12th Annual Conference on Rehabilitation Technology, New Orleans, LA, pp. 194–195.
Swinnen, E. , Lafosse, C. , Van Nieuwenhoven, J. , Ilsbroukx, S. , Beckwee, D. , and Kerckhofs, E. , 2015, “ Neurological Patients and Their Lower Limb Orthotics: An Observational Pilot Study About Acceptance and Satisfaction,” Prosthet. Orthot. Int., 41(1), pp. 41–50.
Wessels, R. , Dijcks, B. , Soede, M. , Gelderblom, G. J. , and De Witte, L. , 2003, “ Non-Use of Provided Assistive Technology Devices, a Literature Overview,” Technol. Disability, 15(4), pp. 231–238.
Celebi, B. , Yalcin, M. , and Patoglu, V. , 2013, “ AssistOn-Knee: A Self-Aligning Knee Exoskeleton,” IEEE/RSJ International Conference on Intelligent Robots and Systems, Tokyo, Japan, Nov. 3–7, pp. 996–1002.
Neckel, N. , Wisman, W. , and Hidler, J. , 2006, “ Limb Alignment and Kinematics Inside a Lokomat Robotic Orthosis,” Annual International Conference of the IEEE Engineering in Medicine and Biology, New York, Aug. 30–Sept. 3, pp. 2698–2701.
Tao, G. D. , 2010, “ Mechanical Bracing Solutions to Decrease Tibial Slippage of Anklebot,” Ph.D., Massachusetts Institute of Technology, Cambridge, MA.
Schorsch, J. F. , Keemink, A. Q. L. , Stienen, A. H. A. , van der Helm, F. C. T. , and Abbink, D. A. , 2014, “ A Novel Self-Aligning Mechanism to Decouple Force and Torques for a Planar Exoskeleton Joint,” Mech. Sci., 5(2), pp. 29–35.
Mirbagheri, M. M. , Niu, X. , Kindig, M. , and Varoqui, D. , 2012, “ The Effects of Locomotor Training With a Robotic-Gait Orthosis (Lokomat) on Neuromuscular Properties in Persons With Chronic SCI,” Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), San Diego, CA, Aug. 28–Sept. 1, pp. 3854–3857.
Meuleman, J. , Van Asseldonk, E. H. F. , and Van Der Kooij, H. , 2013, “ Novel Actuation Design of a Gait Trainer With Shadow Leg Approach,” IEEE International Conference on Rehabilitation Robotics, Seattle, WA, June 24–26, pp. 1–8.
Jezernik, S. , Colombo, G. , Keller, T. , Frueh, H. , and Morari, M. , 2003, “ Robotic Orthosis Lokomat: A Rehabilitation and Research Tool,” Neuromodulation, 6(2), pp. 108–115. [PubMed]
Hocoma, 2017, “ Technical Data Lokomat® Pro,” Volketswil, Switzerland, accessed June 20, 2017, http://www.hocoma.com/fileadmin/user/Dokumente/ Lokomat/TECH_L6_140317_en.pdf
Cyberdyne Inc., 2017, “ What's HAL®?,” Tsukuba, Japan, accessed June 20, 2017, http://www.cyberdyne.jp/english/products/ LowerLimb_medical.html
Guizzo, E. , and Goldstein, H. , 2005, “ The Rise of the Body Bots,” IEEE Spectrum, 42(10), pp. 50–56.
Strickland, E. , 2012, “ Good-Bye, Wheelchair,” IEEE Spectrum, 49(1), pp. 30–32.
Esquenazi, A. , Talaty, M. , Packel, A. , and Saulino, M. , 2012, “ The ReWalk Powered Exoskeleton to Restore Ambulatory Function to Individuals With Thoracic-Level Motor-Complete Spinal Cord Injury,” Am. J. Phys. Med. Rehabil., 91(11), pp. 911–921. [PubMed]
Guerrero, C. R. , Grosu, V. , Grosu, S. , Leu, A. , Ristic-Durrant, D. , Vanderborght, B. , and Lefeber, D. , 2015, “ Torque Control of a Push-Pull Cable Driven Powered Orthosis for the Corbys Platform,” IEEE International Conference on Rehabilitation Robotics (ICORR), Singapore, Aug. 11–14, pp. 25–30.
Lee, Y. , Kim, Y.-J. , Lee, J. , Lee, M. , Choi, B. , Kim, J. , and Park, Y. J. , 2017, “ Biomechanical Design of a Novel Flexible Exoskeleton for Lower Extremities,” IEEE/ASME Trans. Mechatronics, 22(5), pp. 2058–2069.
Lee, Y. , Choi, B. , Lee, J. , Lee, M. , Roh, S.-G. , Kim, J. , Choi, H. , and Kim, Y.-J. , 2016, “ Flexible Sliding Frame for Gait Enhancing Mechatronic System (GEMS),” IEEE 38th Annual International Conference of the Engineering in Medicine and Biology Society (EMBC), Orlando, FL, Aug. 16–20, pp. 598–602.
Kleinjan, J. G. , Dunning, A. G. , and Herder, J. L. , 2014, “ Design of a Compact Actuated Compliant Elbow Joint,” Int. J. Struct. Stab. Dyn., 14(8), p. 1440030.
Machekposhti, F. D. , Tolou, N. , and Herder, J. L. , 2015, “ A Review on Compliant Joints and Rigid-Body Constant Velocity Universal Joints Toward the Design of Compliant Homokinetic Couplings,” ASME J. Mech. Des., 137(3), p. 032301.
Schiele, A. , and Visentin, G. , 2003, “ The ESA Human Arm Exoskeleton for Space Robotics Telepresence,” Seventh International Symposium on Artificial Intelligence, Robotics and Automation in Space, Nara, Japan, May 19–23, pp. 19–23. https://pdfs.semanticscholar.org/86fd/abd1517902b73dc9cd85c172726baafd012f.pdf
Stienen, A. H. A. , Hekman, E. E. G. , van der Helm, F. C. T. , and van der Kooij, H. , 2009, “ Self-Aligning Exoskeleton Axes Through Decoupling of Joint Rotations and Translations,” IEEE Trans. Rob., 25(3), pp. 628–633.
Näf, M. B. , Koopman, A. S. , Baltrusch, S. , Rodriguez-Guerrero, C. , Vanderborght, B. , and Lefeber, D. , 2018, “ Passive Back Support Exoskeleton Improves Range of Motion Using Flexible Beams,” Front. Rob. AI, 5, p. 72.
d'Elia, N. , Vanetti, F. , Cempini, M. , Pasquini, G. , Parri, A. , Rabuffetti, M. , Ferrarin, M. , Lova, R. M. , and Vitiello, N. , 2017, “ Physical Human-Robot Interaction of an Active Pelvis Orthosis: Toward Ergonomic Assessment of Wearable Robots,” J. Neuroeng. Rehabil., 14(1), p. 29. [PubMed]
Cai, V. A. D. , Bidaud, P. , Hayward, V. , Gosselin, F. , and Desailly, E. , 2011, “ Self-Adjusting, Isostatic Exoskeleton for the Human Knee Joint,” Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), Boston, MA, Aug. 30–Sept. 3, pp. 612–618.
Cai, V. A. D. , Bidaud, P. , and Nguyen, V. T. , 2016, “ Control of a Self-Adjusting Lower Limb Exoskeleton for Knee Assistance,” Robot Design, Dynamics And Control (ROMANSY 21), Udine, Italy, June 20–23, pp. 385–392.
Mihelj, M. , Nef, T. , and Riener, R. , 2007, “ ARMin II—7 DoF Rehabilitation Robot: Mechanics and Kinematics,” IEEE International Conference on Robotics and Automation, Roma, Italy, Apr. 10–14, pp. 4120–4125.
Park, H. S. , Ren, Y. , and Zhang, L. Q. , 2008, “ IntelliArm: An Exoskeleton for Diagnosis and Treatment of Patients With Neurological Impairments,” Second Biennial IEEE/RAS-EMBS International Conference on Biomedical Robotics and Biomechatronics (BioRob), Scottsdale, AZ, Oct. 19–22, pp. 109–114.
Otten, A. , Voort, C. , Stienen, A. , Aarts, R. , van Asseldonk, E. , and van der Kooij, H. , 2015, “ LIMPACT:A Hydraulically Powered Self-Aligning Upper Limb Exoskeleton,” IEEE/ASME Trans. Mechatronics, 20(5), pp. 2285–2298.
Beekhuis, J. H. , Westerveld, A. J. , Van Der Kooij, H. , and Stienen, A. H. A. , 2013, “ Design of a Self-Aligning 3-DOF Actuated Exoskeleton for Diagnosis and Training of Wrist and Forearm After Stroke,” IEEE International Conference on Rehabilitation Robotics (ICORR), Seattle, WA, June 24–26, pp. 1–5.
Stormont, T. J. , An, K. N. , Morrey, B. F. , and Chao, E. Y. , 1985, “ Elbow Joint Contact Study: Comparison of Techniques,” J. Biomech., 18(5), pp. 329–336. [PubMed]
Kumar, S. , 2001, “ Theories of Musculoskeletal Injury Causation,” Ergonomics, 44(1), pp. 17–47. [PubMed]
Tuijthof, G. J. M. , and Herder, J. L. , 2000, “ Design, Actuation and Control of an Anthropomorphic Robot Arm,” Mech. Mach. Theory, 35(7), pp. 945–962.
Saccares, L. , Sarakoglou, I. , and Tsagarakis, N. G. , 2016, “ iT-Knee: An Exoskeleton With Ideal Torque Transmission Interface for Ergonomic Power Augmentation,” IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Daejeon, South Korea, Oct. 9–14, pp. 780–786.
Asano, Y. , Mizoguchi, H. , Kozuki, T. , Motegi, Y. , Osada, M. , Urata, J. , Nakanishi, Y. , Okada, K. , and Inaba, M. , 2012, “ Lower Thigh Design of Detailed Musculoskeletal Humanoid 'Kenshiro,” IEEE International Conference on Intelligent Robots and Systems, Vilamoura, Portugal, Oct. 7–12, pp. 4367–4372.
Toxiri, S. , Ortiz, J. , Masood, J. , Fernández, J. , Mateos, L. A. , and Caldwell, D. G. , 2016, “ A Powered Low-Back Exoskeleton for Industrial Handling: Considerations on Controls,” Wearable Robotics: Challenges and Trends, Vol. 16, J. González-Vargas, J. Ibáñez, J. Contreras-Vidal, H. van der Kooij, and J. Pons, eds., Springer, Cham, Switzerland, pp. 287–291.
Toxiri, S. , Ortiz, J. , Masood, J. , Fernandez, J. , Mateos, L. A. , and Caldwell, D. G. , 2015, “ A Wearable Device for Reducing Spinal Loads During Lifting Tasks: Biomechanics and Design Concepts,” IEEE International Conference on Robotics and Biomimetics (ROBIO), Zhuhai, China, Dec. 6–9, pp. 2295–2300.
Vitiello, N. , Lenzi, T. , Roccella, S. , De Rossi, S. M. M. , Cattin, E. , Giovacchini, F. , Vecchi, F. , and Carrozza, M. C. , 2013, “ NEUROExos: A Powered Elbow Exoskeleton for Physical Rehabilitation,” IEEE Trans. Rob., 29(1), pp. 220–235.
Lenzi, T. , Vitiello, N. , De Rossi, S. M. M. , Roccella, S. , Vecchi, F. , and Carrozza, M. C. , 2011, “ NEUROExos: A Variable Impedance Powered Elbow Exoskeleton,” IEEE International Conference on Robotics and Automation, Shanghai, China, May 9–13, pp. 1419–1426.
Cai, V. A. D. , and Bidaud, P. , 2014, “ Self-Adjusting Isostatic Exoskeleton for the Elbow Joint: Mechanical Design Parallel Manipulators,” Advances on Theory and Practice of Robots and Manipulators: Proceedings of ROMANSY 2014 XX CISM-IFToMM Symposium on Theory and Practice of Robots and Manipulators, Springer, Cham, Switzerland, pp. 19–26.
Pierrot, F. , Reynaud, C. , and Fournier, A. , 1990, “ Delta: A Simple and Efficient Parallel Robot,” Robotica, 8(2), pp. 105–109.
Ergin, M. A. , and Patoglu, V. , 2011, “ A Self-Adjusting Knee Exoskeleton for Robot-Assisted Treatment of Knee Injuries,” IEEE International Conference on Intelligent Robots and Systems, San Francisco, CA, Sept. 25–30, pp. 4917–4922.
Beil, J. , and Asfour, T. , 2016, “ New Mechanism for a 3 DOF Exoskeleton Hip Joint With Five Revolute and Two Prismatic Joints,” IEEE RAS and EMBS International Conference on Biomedical Robotics and Biomechatronics, Singapore, June 26–29, pp. 787–792.
Valiente, A. , Herr, H. , and Blanco, E. , 2003, “ Design of a Quasi-Passive Parallel Leg Exoskeleton to Augment Load Carrying for Walking by,” Ph.D. thesis, MIT, Cambridge, MA. https://dspace.mit.edu/handle/1721.1/34412
Shamaei, K. , Cenciarini, M. , Adams, A. A. , Gregorczyk, K. N. , Schiffman, J. M. , and Dollar, A. M. , 2015, “ Biomechanical Effects of Stiffness in Parallel With the Knee Joint During Walking,” IEEE Trans. Biomed. Eng., 62(10), pp. 2389–2401. [PubMed]
Lenzi, T. , Zanotto, D. , Stegall, P. , Carrozza, M. C. , and Agrawal, S. K. , 2012, “ Reducing Muscle Effort in Walking Through Powered Exoskeletons,” Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), San Diego, CA, Aug. 28–Sept. 1, pp. 3926–3929.

Figures

Grahic Jump Location
Fig. 1

Kinematic compatibility can be ensured in different ways: One can attempt to mimic the full joint kinematics or rely on simplified mechanisms. To mimic the full joint kinematics is often difficult and therefore not often practiced. This is mostly because the exact location of the human joint rotation axis must be known and replicated in an external structure. Simplified structures that avoid any relative movement between the exoskeleton and the human are simpler to implement, since they do not require that knowledge. Nonanthropomorphic exoskeletons and end effector-based devices usually avoid aligning multiple joints, but suffer from a number of disadvantages compared to anthropomorphic exoskeletons. Finally, anthropomorphic exoskeletons can be aligned manually, by the use of compliant elements or by adding kinematic redundancy. Anthropomorphic exoskeletons, which use any of these alignment techniques, are said to implement misalignment compensation.

Grahic Jump Location
Fig. 2

Schematic representation of a 1DOF human limb and an exoskeleton, where the joint axes are not aligned: (a) A change of angle α in both the exoskeleton and the human leads to a radial displacement dl and a tangential displacement dw and (b) if the human limb is connected by a brace, an angular change α will lead to a radial displacement L and a small angular rotation γ. Figure based on Ref. [35].

Grahic Jump Location
Fig. 3

Overview of the different misalignment concepts based on adding kinematic redundancy: (a) two revolute joints in series with a prismatic joint, revolute-revolute-prismatic (RRP) joint, (b) one revolute joint in series with two prismatic joints, revolute-prismatic-prismatic (RPP) joint, (c) three revolute joints in series, revolute-revolute-revolute (RRR) joint. In the subfigures (a) and (c), the rotation axes are depicted as parallel, but arbitrary angles between them are possible. Similarly, in subfigure (b), the two prismatic joints can be orthogonal, instead of parallel, as depicted here.

Grahic Jump Location
Fig. 4

The use of anthropomorphic assistive exoskeletons requires the introduction of misalignment compensation techniques to ensure safe and high-performance interaction with the user. These techniques can be divided into three main groups: manual alignment, use of compliant elements, and addition of kinematic redundancy. For two of these, a further subdivision is possible. When compliant elements are used, they can be introduced either at the brace level or directly in the frame. Additionally, it is also possible to use compliant exoskeleton joints. As for kinematic redundancy techniques, kinematic compatibility is typically ensured by adding three additional DOFs to the device at a time, consisting of revolute and prismatic joints. Depending on the type of the additional elements, a division can be made on three categories: revolute-revolute-prismatic (RRP), revolute-prismatic-prismatic (RPP), and revolute-revolute-revolute (RRR).

Grahic Jump Location
Fig. 5

Hip module of the S-assist exoskeleton: The vertical sliding mechanism compensates for the movement between the device and it is wearer due to the misalignment of the ab-/adduction axis. Picture adapted with permission from Ref. [84].

Grahic Jump Location
Fig. 6

Flexible frame used in exoskeleton by Lee et al. [84,85]: These structure were designed in such a way that they are rigid in the direction where torque needs to be transmitted to the user, but are flexible in other directions (gray arrow F). Therefore, they can be tightly fitted to each individual wearer and slightly deform to alleviate misalignment effects. Picture adapted with permission from Ref. [85].

Grahic Jump Location
Fig. 7

Dimensioning properly slender beams, flexures, revolute joints (a) and spherical joints (b) can be obtained easily. Figures based on Ref. [86].

Grahic Jump Location
Fig. 8

Misalignment between an exoskeleton joint and a human joint: (a) the exoskeleton is horizontally misaligned by a distance x and vertically misaligned by a distance y. If a torque TAct is applied in absence of any compensation mechanism, the brace is displaced by a distance L and rotated by a small angle γ [35]. (b) If a misalignment compensation mechanism in the form of a revolute-revolute-prismatic (RRP) joint configuration is introduced, the brace does not move; instead, the displacement L is compensated with the prismatic joint and the small rotation γ is compensated by the additional revolute joint. Figures based on Ref. [35].

Grahic Jump Location
Fig. 9

Passive back support exoskeleton: In the elastic spinal module, two rotational DOFs are implemented with the elasticity of carbon fiber beams hinge joint (passively actuated). The misalignment compensation at the hip joint as well as a fitting mechanism is realized with a combination of an RRR mechanism (highlighted with arrows) and a prismatic joint. Pictures adapted with permission from Ref. [90].

Grahic Jump Location
Fig. 10

Representation of the kinematic chain of the hip exoskeleton by D'Elai et al. joints (1)–(4) are passive DoFs. Joints (D)–(F) represent the internal/external, the ab/adduction and, the flexion/extension joint, respectively. The pelvis is loaded through its interface (A). Picture adapted with permission from Ref. [91].

Grahic Jump Location
Fig. 11

RPP misalignment compensation mechanism exploited in different exoskeleton: (a) lower limb exoskeleton for knee assistance: only the knee joint is actuated in this exoskeleton; all the other joints serve to compensate for misalignment. For instance for the leg and knee flexion/extension, an RPP strategy (highlighted with arrows) is implemented below the knee. (b) Bartenbach exoskeleton: the RPP mechanism is used to compensate for the misalignment that might take place between the leg joint axes and the corresponding exoskeleton ones. It is worth noting that RPP mechanisms are used (i) to compensate for the misalignment between the exoskeleton rotational joints (allow circle) and the human ones and, (ii) to adjust the position of the cuffs on the leg (orthogonal prismatic joints—highlighted with arrows). Pictures adapted with permission from Refs. [11] and [93].

Grahic Jump Location
Fig. 12

(a) A small misalignment δ between an exoskeleton joint and the human axis of rotation leads to a reaction force R1 in the exoskeleton and an equal and opposite undesired reaction force R2 in the human joint, when a torque TAct is applied on the human limb. (b) By adding two linear DOFs to the exoskeleton, the reaction forces are absorbed by movements of the sliders.

Grahic Jump Location
Fig. 13

(a) Kinematic structure of ARMinII rehabilitation robot: joint axes 1–3 are close to the shoulder rotational axes while two passive and vertical translational degrees-of-freedom (squared joints) allow for misalignment compensation of the shoulder, rendering this an RPP mechanism. Figure based on Ref. [94]. (b) Kinematic structure of IntelliArm: squared joints constitute a Cartesian stage allowing the robot to compensate for the motion of the shoulder girdle. Figure based on Ref. [95].

Grahic Jump Location
Fig. 14

(a) Schematic representation of the RRR mechanism applied on a wrist. The RRR mechanism is implemented via three rotational joints labeled, respectively, with numbers 1, 2, and 3. Three configurations are represented: neutral configuration, wrist abduction, and wrist adduction. The mechanism accommodates for misalignments with two orthogonal translational movements, from x (y) until x + Δx (y + Δy). One of these joints can be actuated directly or remotely by cable systems. Figure based on Ref. [88]. (b) Wrist module for autonomous use or as an expansion to the Dampace/LIMPACT exoskeleton: A change in the adduction angle ΘAD or abduction angle ΘAB is preserved by the kinematic structure. Picture adapted with permission from Ref. [97].

Grahic Jump Location
Fig. 15

(a) Elbow lifting task with the original exoskeleton. An external load is applied at point C, producing a torque Tjoint around the human elbow joint (A). The serial 4-bar mechanism, which is connected to the lower arm at point B and anchored at point D, has to produce a torque T4–bar to compensate for the external load. (b) Addition of a static balancing mechanism in form of springs onto the original design. (c) Mechanical execution of the improved elbow joint. Pictures adapted with permission from Ref. [74].

Grahic Jump Location
Fig. 16

Left: Kinematic structure of the Robomate exoskeleton: only the flexion rotational joints are actuated; ball and socket joints are represented with spheres. The RRR mechanism that compensates for the abduction misalignment is highlighted with arrows. Right: RoboMate exoskeleton to reduce spinal load during lifting tasks. Pictures adapted with permission from Refs. [103] and [104].

Grahic Jump Location
Fig. 17

(a) Left: ASSISTON-Knee exoskeleton, worn by a human subject; right: detail of the inner RRR-mechanism of the device highlighted with circle. (b) CAD drawing and physical model of the three-RRP mechanism. Pictures adapted with permission from Refs. [71] and [109].

Grahic Jump Location
Fig. 18

Hip exoskeleton by Beil et al. [110], which accounts for all 3DOF of the hip. L2 is the reference frame related to the abduction/adduction joint, which needs to be manually aligned by acting on clampers. L8 represents the reference frame for the flexion/extension joint. L3–L7 reference frames are related to joints that compensate for misalignment, stemming from internal external rotation. Reference frames circled with continuous line refer to rotational joints; reference frames circled with dashed line refer to prismatic joints. Picture adapted with permission from Ref. [110].

Grahic Jump Location
Fig. 19

A cam mechanism is used to compensate for misalignment of the hip ab-/adduction axis. Pictures adapted with permission from Ref. [28].

Tables

Table Grahic Jump Location
Table 1 Classification of misalignment compensation mechanisms for state-of-the-art exoskeletons, according to misalignment compensation strategy. Each strategy is evaluated by previously defined requirements. Each requirement is scored by using one of the following symbols: “−−,” “−,” “+” and “++.” In the order given, they stand for: “not good at all,” “rather bad,” “ok” and “very good,” respectively.

Errata

Some tools below are only available to our subscribers or users with an online account.