Energy Shaping Control Of Powered Lower Limb Exoskeletons For Assistance Of Human Locomotion

The majority of powered lower-limb exoskeletons nowadays are designed to rigidly track time-based kinematic patterns, which forces users to follow specific joint positions. This kinematic control approach is limited to replicating the normative joint kinematics associated with one specific task and user at a time. These pre-defined trajectories cannot adjust to continuously varying activities or changes in user behavior associated with learning during gait rehabilitation. Time-based kinematic control approach must also recognize the user’s intent to transition from one task-specific controller to another, which is susceptible to errors in intent recognition and does not allow for a continuous range of activities. Moreover, fixed joint patterns also do not facilitate active learning during gait rehabilitation. People with partial or full volitional control of their lower extremities should be allowed to adjust their joint kinematics during the learning process based on corrections from the therapist. To address this issue, we propose that instead of tracking reference kinematic patterns, kinetic goals (for example, energy or force) can be enforced to provide a flexible learning environment and allow the user to choose their own kinematic patterns for different locomotor tasks. In this dissertation, we focus on an energetic control approach that shapes the Lagrangian of the human body and exoskeleton in closed loop. This energetic control approach, known as energy shaping, controls the system energy to a specific analytical function of the system state in order to induce different dynamics via the Euler-Lagrange equations. By explicitly modeling holonomic contact constraints in the dynamics, we transform the conventional Lagrangian dynamics into the equivalent constrained dynamics that have fewer (or possibly zero) unactuated coordinates. Based on these constrained dynamics, the matching conditions, which determine what energetic properties of the human body can be shaped, become easier to satisfy. By satisfying matching conditions for human-robot systems with arbitrary system dimension and degrees of actuation, we are therefore able to present a complete theoretical framework for underactuated energy shaping that incorporates both environmental and human interaction. Simulation results on a human-like biped model and experimental results with able-bodied subjects across a variety of locomotor tasks have demonstrated the potential clinical benefits of the proposed control approach.

Lower Limb Exoskeleton, a wearable robot that is designed to provide lower limb assistance to users, has been rapidly developed in the previous decade. The goal of these robots is to replace human labor with robots while still having humans involved. However, while these robot suits provide sufficient assistance to the users, the efficiency of the robot is often overseen. Thus, restrict the exoskeleton's operating time or required it to connect to an external power supply. However, there is plenty of energy wasted in human motions. In this study, we target "loaded bipedal walking" as the primary motion to assist. In chapter 2, we applied trajectory optimization on different mechanical designs for lower-limb exoskeletons. It is commonly known that humans tend to use more energy to walk compared to other limb-based locomotion animals. This higher energy usage is due to "heel strikes" and "negative work" during human gait. Passive walkers elevate this phenomenon by utilizing elastic joints that absorb/reuse some of the negative work. The objective of this study is to absorb energy at one phase of the gait cycle, store it, and then release it at a later phase through the use of a lower limb exoskeleton. Knee geometry is one important factor in energy efficiency during gait. Animals with reversed knees compared to humans (backward knee), such as ostriches, exhibit improved energy efficiency. As part of this study, new energy optimization strategies were developed utilizing collision-based ground reaction forces and a discrete lagrangian. The minimal cost of transport (CoT) gait patterns were calculated with both forward-knee and backward-knee human-exoskeleton models. Simulation results show that wearing a backward-knee exoskeleton can reduce the CoT by 15% of while carrying external loads ranging from 20 to 60 kg. In addition, when the exoskeleton utilized energy recycling, the CoT was shown to be further reduced to 35%. These simulation results suggested that the optimal design for an exoskeleton aimed at utilizing energy recycling principles should incorporate backward-knee configurations much like those found in energy-efficient biped/quadruped animals. In fact, since the potential energy sources (heel strikes, negative work) and the main energy consumer (ankle push-off) occurs in the opposite legs, the ideal actuators for the exoskeleton need to be able to recycle, store, and transfer energy between different legs. To satisfy the actuator's requirements from chapter 2, in chapter 3 we choose pneumatic actuators as the actuator for our exoskeleton. Pneumatic actuators are a popular choice for wearable robotics due to their high force-to-weight ratio and natural compliance, which allows them to absorb and reuse wasted energy during movement. However, traditional pneumatic control is energy inefficient and difficult to precisely control due to nonlinear dynamics, latency, and the challenge of quantifying mechanical properties. To address these issues, In chapter 3, we developed a wearable pneumatic actuator with energy recycling capabilities and applied the sparse identification of nonlinear dynamics (SINDy) algorithm to generate a nonlinear delayed differential model from simple pressure measurements. Using only basic knowledge of thermal dynamics, SINDy was able to train models of solenoid valve-based pneumatic systems with a training accuracy of 90.58% and a test accuracy of 86.44%. The generated model, when integrated with model predictive control (MPC), resulted in a 5% error in pressure control. By using MPC for human assistive impedance control, the actuator was able to output the desired force profile and recycle around 88% of the energy used in negative work. These results demonstrate an energy-efficient and easily calibrated actuation scheme for designing assistive devices such as exoskeletons and orthoses. In chapter 4, we presented Pneumatic Exoskeleton with Reversible Knee (PERK). It utilizes the pneumatic actuators we developed in chapter 3 and the control strategies we concluded in chapter 2. Three clinical trials were done on three different test subjects. The results showed despite different walking patterns across different test subjects, there is less potential energy change during the swing phase of walking, potentially reducing the energy loss during the heel strike. In addition, during the double support phase, there is less energy consumption in the pneumatic system while configuring it as backward-knee, indicating it is easier or more intuitive for the user to have the exoskeleton recycling the dissipated energy with the backward-knee mechanism.

Mots-clés de l'auteur: Human locomotion ; particle swarm algorithm (PSO) ; neuromuscular model ; reflex ; central pattern generator (CPG) ; spinal cord injury ; paraplegia ; rehabilitation ; exoskeleton ; orthese.

The field of exoskeletons has undergone an evolution from complex full-body exoskeletons that did not (yet) produce the expected results towards simpler single-joint exoskeletons that can improve the mobility of people. While full-body and single-joint exoskeletons certainly have appropriate applications, we need to get a better understanding of the distal and proximal assistive mechanisms and provide insights that are currently lacking on how to assist walking in an even simpler way than single-joint exoskeletons. This dissertation details an iterative approach toward the development of simplified and efficient assistance strategies for improving human locomotion. We first conducted an experiment to observe the human response to proximal and distal perturbations by altering the treadmill grade and footwear inclination. The results indicate that the metabolic rate is predominantly sensitive to changes in the center of mass (COM) mechanics and further motivate the development of devices that can assist walking at the level of the COM. We then developed a robotic tether system that allows applying desired cyclic force profiles as a function of step time to provide whole-body assistance during walking. By leveraging the system capabilities, we performed an experiment and simple pendulum simulation to investigate the effects of timing and magnitude for non-constant force profiles at the COM. Through these experiments and the simulation, we found that assistance at the COM during the double stance phase can efficiently reduce the metabolic rate of walking half. Surprisingly, assisting propulsion did not maximize the reduction in metabolic rate, and our pendulum model revealed that the reduction in metabolic rate can instead be explained by the assistance of COM acceleration at the beginning of the step. Ultimately, our long term goal is to develop similar strategies to populations with gait disabilities, but as a primary step, we investigated the biomechanical mechanisms to assist lower limb joints using timed forward forces at the COM. To that end, we assessed the underlying mechanisms of muscle and joint parameters that explain the effects of timing and magnitude of horizontal forces at the COM on metabolic rate. The results show that the metabolically optimal timing assisted the ankle muscles that are responsible for push-off, and the knee and hip muscles that are responsible for collision. Based on these findings, it seems possible to assist different joints by different amounts by varying the timing of forces at the COM. This could be useful in clinical populations for providing ‘targeted’ joint-specific assistance without having to switch between different exoskeletons. We expect our experimental findings to provide knowledge on optimal force profiles that could be used for treadmill exercise therapy, motorized ‘rollator’-style assistive devices for walking, and even it could even inspire new strategies for combined actions of the ankle, knee, and hip of full-body exoskeletons. Timed forces at the COM could be used to assist patients with impaired gait and facilitate proactive user participation that has been identified as a critical factor in improving locomotor outcomes for rehabilitation robotics.

Lower extremity powered exoskeletons (LEPE) are powered orthoses that enable persons with spinal cord injury (SCI) to ambulate independently. Since locomotor therapy must be specific and resemble natural gait patterns, to promote motor recovery, current LEPE control architectures may be inappropriate since they typically use able-bodied, pre-recorded reference position and force data, at normal walking speeds, to define exoskeleton motion and predict torque assistance. This thesis explored two aspects: a) able-bodied walking dynamics between 0.2 m/s and the person's self-paced speed to provide a biomimetic basis for LEPE control and b) musculoskeletal modelling of LEPE-human dynamics. For walking dynamics, appropriate regression equations were developed for stride, kinematic, and kinetic parameters. These equations can be used by LEPE designers when constructing angular trajectories and forces for LEPE control at any given speed. An inflection point at 0.5 m/s was identified for temporal stride parameters; therefore, different walking strategies should be considered for walking above and below this point. The full body musculoskeletal model (Anybody) of persons with SCI using the ARKE LEPE incorporated all external contact forces and inertial properties (exoskeleton and person) and was driven using real LEPE SCI user kinematics and kinetics. For the lower extremity, large dorsiflexion range of motion, large device anterior tilt, incomplete knee extension, and uncontrolled center of pressure forward progression lifted the heel during stance. This triggered step termination before trajectory tracking at the knee and hip was complete, thereby reducing hip extension, increasing knee flexion through stance, increasing knee and hip support moments, and increasing thigh and shank strap reaction forces. This also shortened effective participant limb length, further shortening step-length and LEPE walking speed. For the upper-limbs, LEPE users walked with more anterior trunk tilt and twice the shoulder flexion angle, compared with persons with incomplete SCI. This increased forces and moments at the crutch, shoulder, and elbow. Crutch floor contact periods were 30-40% longer, resulting in upper-extremity joint impulses 5 to 12 times greater than previously reported. Improved step-completion and upright posture would reduce support loads on the crutches and upper-limbs, and would further improve LEPE-human lower limb interaction forces. Improved upright posture and LEPE-human interaction forces would enhance mobility and quality of movement for people with SCI.

Locomotion is inherently an energy regulation challenge; ground impacts deplete the mechanical energy of the walking system with every step. When a person’s leg is amputated, one of the conventional medical devices used to help them recover their mobility is a passive prosthesis. However, this device is incapable of doing positive work on the human body to counteract energy depletion and restore the user’s mechanical energy. Powered prostheses have been developed and researched to address this, but recent control methods have focused on tracking joint trajectories or impedance while ignoring the fundamental kinetic aspect of human locomotion. The prevailing goal of this work is to construct a control method for a powered lower-limb prosthesis that explicitly and directly enhances the kinetics of the combined human-prosthesis system to assist human locomotion. The method proposed to accomplish this utilizes energy and passivity based control techniques to modify the dynamics of the prosthesis. This dissertation develops control theory related to these techniques for autonomous bipedal robots so that they can then be translated onto the target prosthesis system. Specifically, it shows how to use energy shaping and regulation to change characteristics of a walking gait, like walking speed, via switching of a small set of physically meaningful parameters. Experimental results that demonstrate proof-of-concept on a powered knee-ankle prosthetic leg are presented.

For over a century, technologists have strived to develop autonomous leg exoskeletons that reduce the metabolic energy consumed when humans walk and run, but such technologies have traditionally remained unachievable. In this thesis, I present the Augmentation Factor, a simple model that predicts the metabolic impact of lower limb exoskeletons during walking. The Augmentation Factor balances the benefits of positive exoskeletal mechanical power with the costs of mechanical power dissipation and added limb mass. These insights were used to design and develop an autonomous powered ankle exoskeleton. A lightweight electric actuator mounted on the lower-leg provides mechanical assistance to the ankle during powered plantar flexion. Use of the exoskeleton significantly reduced the metabolic cost of walking by 11 ± 4% (p = 0.019) compared to walking without the device. In a separate study, use of the exoskeleton reduced the metabolic cost of walking with a 23 kg weighted vest by 8 ± 3% (p = 0.012). A biomechanical study revealed that the powered ankle exoskeleton does not simply replace ankle function, but augments the biological ankle while assisting the knee and hip. Use of the powered ankle exoskeleton was shown to significantly reduced the mean positive power of the biological ankle by 0.033 ± 0.006 W/kg (p

In the development of powered lower-limb prostheses, providing sufficient power and torque to support amputees' locomotion is a major challenge, considering prostheses' weight and size limits. Furthermore, regulating the power delivery during locomotion is equally important that gives amputees safe and natural movements. This dissertation aims to address these challenges by investigating new approaches in the actuation and control of powered lower-limb prostheses, with the overarching objective to obtain compact, powerful lower-limb prostheses that interact with amputees and the environment in a coordinated manner. The initial efforts were focused on the design and control of transfemoral (TF, also known as above-knee) prostheses powered by pneumatic muscles, an extraordinary actuator with superb power-to-weight ratio. The first prototype incorporates powered knee and ankle joints in a volumetric profile similar to that of human leg. The unique feature is a single-acting-spring-return mechanism, in which a single pneumatic muscle drives the motion in the torque-demanding direction, while a set of mechanical springs drives the motion in the opposite direction. A finite-state impedance controller has been developed for this prosthesis, which was demonstrated to provide a natural gait. Based on previous success, a novel type of pneumatic muscle, namely double-acting sleeve muscle (DASM), was examined to replace traditional pneumatic muscle. Incorporating a second chamber, the DASM is able to provide additional extensional force without using return springs. Therefore, the prosthesis can be significantly simplified into a more compact and lightweight device. Compared with pneumatic muscles, traditional cylinder-type actuators are more technologically mature. Therefore, the subsequent efforts were to develop a pneumatic cylinder-actuated TF prosthesis, which has great potential for real-world applications. All peripheral components were integrated, including a carbon fiber air tank as the energy source, and the prosthesis' capability of independent, untethered operation was demonstrated in human walking test. In addition to the improvement of prosthetic design, control methods were also investigated. The results include an integrated walking -- stair climbing controller and a sit-to-stand controller. Both were developed based on biomechanical analysis of the knee dynamics in human locomotion. In the walking -- stair climbing control system, an improved finite state impedance controller was constructed, which incorporates a unique time function to enable gradual energy injection during weight acceptance phase. An intuitive thigh position-based switching condition was introduced to merge the walking and stair climbing controllers into one system. In the sit-to-stand controller, a similar controller was established, which eliminates the need for a state machine and significantly simplifies the controller tuning and implementation. The human testing was conducted with results demonstrating the effectiveness of both control systems.