Development of a Compliant Gait Rehabilitation Exoskeleton
Pneumatic muscle actuated rehabilitation robotic devices have been widely researched, because of its intrinsic compliance and high power to weight ratio[19, 78, 149]. Task specific gait rehabilitation training imposes strict torque, range of motion and bandwidth requirements to the robotic exoskeleton design. However, the PM’s nonlinear and hysteresis behavior, slow pressure dynamics and negative correlation between its force output and contracting length make the development even more challenging. To address such challenges, a new robotic GAit Rehabilitation EXoskeleton (GAREX) has been developed in order to facilitate task specific gait rehabilitation with controlled intrinsic compliance. GAREX is developed with the potential clinical experiments in mind. Detailed actuation system design analysis is carried out to meet the requirements of task specific robotic gait training. Implementations on mechanical, electronic and software aspects ensure the safety of the training subject, which is of paramount importance. GAREX also has a modular construction to accommodate anthropometrics of most of the population.
An exoskeleton type robot with low inertial moving parts, powerful and compliance actuators would provide a good robotic platform for AAN gait training. Exoskeleton type of gait rehabilitation robots like LokoMat  and Active Leg Exoskeleton (ALEX)  have geared electric motors or linear motors to provide linear actuation. The linear actuations together with levers produce driving torques to the actuated joints. To some exoskeletons, DC electric motor and gear units are directly mounted to the actuated joints to provide required rotational movements [65, 77]. To meet the torque requirement of gait rehabilitation training, the adopted motor and gear reduction units are usually heavy in weight. Thus, the inertia of the lower limb exoskeleton is significantly increased. As a result, some control strategies such as impedance control may not be easy to implement. Being heavily geared also means high endpoint impedance, high stiffness and non-backdrivable . However, stiff actuators could generate large force in response the patient’s undesirable movements or impacts and further lead to discomfort or even safety problems . One approach to reduce the inertia is to detach the motor units from the moving segment of the exoskeleton. Cable systems are often utilized as media between the distally located actuators and the actuated joints . Series elastic actuators  provide a solution to the high endpoint impedance meanwhile adding compliance and back-drivability to the actuated exoskeleton joints. LOPES  was developed at the University of Twente with both Series elastic actuation (SEA) and Bowden cable power transmission. Taking the advantages of the combination, both high and low impedance control strategies have been implemented to achieve robot guidance or near transparent patient in charge walking. However, spring stiffness of the SEA system is limited; thus, the extent of high impedance that can be reached is restricted by the mechanical design. Moreover, the utilization of Bowden cable could also affect the control precision and bandwidth, reduce power transfer efficiency, and be prone to wearing [19, 79]. Pneumatic muscle (PM) actuators provides a good option to reduce exoskeleton inertia as well as achieve variable actuation stiffness, due to the actuators’ high power to weight ratio and intrinsic compliance . With such properties, PMs have been employed in various rehabilitation robotic applications [19, 78, 153]. To achieve effective robotic gait rehabilitation, the training is required to be as task specific as possible . Hence, the robots need to assist or guide the subjects to walk in or similar to normal gait speed and pattern. Transferring this to the robot development, the following specifications are to be satisfied: (1) multiple actuated DoFs to reproduce gait pattern; (2) sufficient range of motion for the powered joints; (3) sufficient joint actuation torque to guide severely impaired subjects during training; (4) adequate controlled bandwidth to facilitate the training. Moreover, the intrinsic compliance property is one of the main motivations of adapting PMs in rehabilitation robotics. It is equally important to design the actuation system that opens the possibilities for the researcher to control the compliance of the exoskeleton during gait training. In the past, gait rehabilitation exoskeletons actuated by PMs have been developed. Beyl et al. [153, 155] developed a lower limb exoskeleton actuated by pleated pneumatic artificial muscle. Both torque and trajectory control approaches were investigated. However, with the only actuated DoF at the knee joint, the device’s potential use in clinical trials is limited. Hussain et al.  developed a treadmill based robotic gait rehabilitation exoskeleton with two both the hip and knee joints actuated in the sagittal plane. Only the maximum joint torques of the exoskeleton were mentioned instead of detailed torque analysis which may not be sufficient to justify the torque requirement of task specific gait rehabilitation had been met. Moreover, safety and anthropometric adaptability factors about the exoskeleton design were not reported in details too. There are needs to develop a new PM actuated gait rehabilitation robotic exoskeleton with the requirements of task specific training in mind. All the above factors determining the feasibility of applying the gait rehabilitation robots in clinical setting should be thoroughly investigated. This chapter will firstly briefly describe the support structure and trunk mechanism of a new GAit Rehabilitation Exoskeleton (GAREX). This will be followed by actuation system and mechanism design of the unilateral lower limb exoskeleton. Pneumatic system and instrumentations will be presented next. Lastly, the safety anthropometric adaptability design of the GAREX system will be covered.
Support structure and trunk mechanism
GARES consists of three major mechanical modules: the support structure, the trunk mechanism, and most importantly the unilateral low limb exoskeleton which are illustrated. The support structure is constructed with MiniTec frames. All the mechanical, pneumatic and electronic components are mounted to the structure. The structure is usually rigidly connected to the rehabilitation treadmill for treadmill based gait rehabilitation experiments. It also has four wheels, so over-ground training with GAREX is also made possible. Due to the modular mechanical design, the trunk mechanism’s vertical and horizontal position relative to the support structure can be easily adjusted to suit various patients or switch between treadmill based and over-ground gait training.
Lower Limb Exoskeleton
The main module of the robotic platform is the lower limb exoskeleton (LLE) which has a unilateral design. In order to help patients to reproduce gait patterns, the sagittal plane rotations of the LLE’s hip and knee joints are actuated by PMs, taking the advantages offered by such actuators discussed in the Introduction section of this chapter. The major challenge of LLE design is also from the PM actuation system. The main design criteria of the LLE are listed as follow:
– The capability of providing sufficient assisting torque and ROM to facilitate robotic gait rehabilitation.
– Compact and light weight design that fit the anthropometric data of majority of the population.
– Robust enough for further experiments with human subjects and potential clinical studies.
Actuation system of the lower limb exoskeleton
Before getting into the design process, it is worthwhile to briefly mention the operation principles of the PM actuators. PMs tend to contract and apply uni-directional pulling actuation to attached objects, when inflated. Hence, it is a common practice to adopt antagonistic configurations of PMs to actuate rotational joints for bi-directional actuation [158-160]. The antagonistic PMs of the exoskeleton are comparable to the extensor and flexor muscle groups of the human hip and knee joints. Therefore, regarding to the torque requirement of a joint, the extension and flexion PMs were analysed separately. The force generated by a PM actuator is positively correlated to its diameter and inflating pressure and negatively correlated to the contracting percentage with respect to its original length. In other word, there is a trade-off between the ROM and maximum torque at the extreme positions that a certain PM system can provide. To meet the first design criterion, PMs need to be longer and thicker in order to provide more torque and greater ROM. However, such design could potentially violate the second criterion, because the segment of the exoskeleton cannot be too much longer than the corresponding segment of the human lower limb. Hence, the design problem was simplified into determining the quantity, type, diameter and length of PM actuators for the actuated joints, as well as the moment arms of the antagonistic PMs, so first two criteria mentioned previously could both be satisfied. The PMs manufactured by FESTO were selected. Compared to in-house manufactured PMs, commercially available PMs would provide better reliability, reduce the inter-PM variance and hence reduce the uncertainty in modelling. The FESTO PMs can be made in three diameters 10, 20 and 40 millimetres and their length can be customized, which provides good design flexibility. In the same situation (pressure, original PM length and contraction percentage), the 40 mm diameter PMs would provide pulling force and hence joint torque. However, due to their bulkiness, mounting them to the exoskeleton would result much longer and thicker exoskeleton segment than actual human lower limb segments. One solution to this problem is to have Bowden cables as transmission and mount the PMs at a remote location . Bowden cable power transmission however conflicts with the one of the initial motivations for employing PMs, which is being light for direct exoskeleton attaching. Nonetheless, the Bowden cable would add modelling uncertainties and reliability issues.
Mechanical design of the lower limb exoskeleton
After selecting the pneumatic actuators, the following step was to design and construct the LLE that integrates with the antagonistic PM actuators. The LLE also has a modular design. The knee and hip modules each contains its corresponding joint actuation system. The two modules are connected to form the thigh segment of the LLE. The relative position of the two modules can be adjusted to ensure the hip and knee joints of the LLE are aligned with the subject’s joints during an experiment. Each of the modules has a rigid leg brace attached. During training, the subject is strapped to the braces via straps. The designs by Hussain et al.  and Choi et al. has PMs located at both the thigh and shank segments of the exoskeleton for actuation of the hip and knee joints respectively. On GAREX, all 8 PM actuators were placed to the thigh segment of LLE. Such design transfers the actuators’ inertia from the shank to the thigh segment, and thus reduces the torque requirements of both the hip and knee joints for producing the same movement. In order to accomplish a robust working prototype in a short time frame, most of the parts of the LLE were made of laser cut aluminum plates. Some key structure parts, such as the shafts and bearing houses of the actuated joints, were machined steel parts.
1.1 Background and motivation
1.2 Research objectives
1.3 Thesis outline and contributions
2. Literature review
2.1 Conventional gait rehabilitation approaches
2.2 Gait rehabilitation robots
2.3 Control strategies of robotic gait rehabilitation
2.4 Clinical effectiveness of robotic gait rehabilitation training
2.5 Approaches to assess gait ability
3. Development of a Compliant Gait Rehabilitation Exoskeleton
3.2 Support structure and trunk mechanis
3.3 Lower Limb Exoskeleton
3.5 Safety and adaptability of GAREX
3.6 Discussion and conclusions
4. Force Dynamics Model of the Pneumatic Muscle Actuators
4.2 Modelling the dynamic characteristics of the PM
4.3 Experimental setup and procedures
4.4 Result Analysis
4.5 Discussion and conclusions
5. Joint Space Trajectory and Compliance Control
5.2 System modelling
5.3 SISO trajectory control of the knee joint mechanism
5.4 MIMO sliding mode trajectory and compliance controller
5.5 Discussion and conclusions
6. MIMO Sliding Mode Control System for GAREX
6.2 Reference gait trajectory generation
6.3 MIMO Sliding mode controllers for GAREX
6.4 From average antagonistic PM pressure to joint compliance
6.5 Experimental validation of the MIMO SM control system of GAREX
6.6 Discussion and conclusions
7. FLCA Based Assist-as-needed Gait Rehabilitation for GAREX
7.2 The assessment of active participation
7.3 Implementation of the FLCA controller
7.4 Experimental validation
7.5 Discussion and conclusions
8.1 Impact and contributions
8.2 Outlook and future work
GET THE COMPLETE PROJECT