Adaptive Control of Exoskeleton Robots for Periodic Assistive Behaviours Based on EMG Feedback Minimisation

In this paper we propose an exoskeleton control method for adaptive learning of assistive joint torque profiles in periodic tasks. We use human muscle activity as feedback to adapt the assistive joint torque behaviour in a way that the muscle activity is minimised. The user can then relax while the exoskeleton takes over the task execution. If the task is altered and the existing assistive behaviour becomes inadequate, the exoskeleton gradually adapts to the new task execution so that the increased muscle activity caused by the new desired task can be reduced. The advantage of the proposed method is that it does not require biomechanical or dynamical models. Our proposed learning system uses Dynamical Movement Primitives (DMPs) as a trajectory generator and parameters of DMPs are modulated using Locally Weighted Regression. Then, the learning system is combined with adaptive oscillators that determine the phase and frequency of motion according to measured Electromyography (EMG) signals. We tested the method with real robot experiments where subjects wearing an elbow exoskeleton had to move an object of an unknown mass according to a predefined reference motion. We further evaluated the proposed approach on a whole-arm exoskeleton to show that it is able to adaptively derive assistive torques even for multiple-joint motion.     

Development and Analysis of an Electrically Actuated Lower Extremity Assistive Exoskeleton

An electrically actuated lower extremity exoskeleton is developed,in which only the knee joint is actuated actively while other joints linked by elastic elements are actuated passively.This paper describes the critical design criteria and presents the process of design and calculation of the actuation system.A flexible physical Human-Robot-Interaction (pHRI) measurement device is designed and applied to detect the human movement,which comprises two force sensors and two gasbags attached to the inner surface of the connection cuff.An online adaptive pHRI minimization control strategy is proposed and implemented to drive the robotic exoskeleton system to follow the motion trajectory of human limb.The measured pHRI information is fused by the Variance Weighted Average (VWA) method.The Mean Square Values (MSV) of pHRI and control torque are utilized to evaluate the performance of the exoskeleton.To improve the comfort level and reduce energy consumption,the gravity compensation is taken into consideration when the control law is designed.Finally,practical experiments are performed on healthy users.Experimental results show that the proposed system can assist people to walk and the outlined control strategy is valid and effective.     

Muscle recruitment and coordination with an ankle exoskeleton

Exoskeletons have the potential to assist and augment human performance. Understanding how users adapt their movement and neuromuscular control in response to external assistance is important to inform the design of these devices. The aim of this research was to evaluate changes in muscle recruitment and coordination for ten unimpaired individuals walking with an ankle exoskeleton. We evaluated changes in the activity of individual muscles, cocontraction levels, and synergistic patterns of muscle coordination with increasing exoskeleton work and torque. Participants were able to selectively reduce activity of the ankle plantarflexors with increasing exoskeleton assistance. Increasing exoskeleton net work resulted in greater reductions in muscle activity than increasing exoskeleton torque. Patterns of muscle coordination were not restricted or constrained to synergistic patterns observed during unassisted walking. While three synergies could describe nearly 95% of the variance in electromyography data during unassisted walking, these same synergies could describe only 85–90% of the variance in muscle activity while walking with the exoskeleton. Synergies calculated with the exoskeleton demonstrated greater changes in synergy weights with increasing exoskeleton work versus greater changes in synergy activations with increasing exoskeleton torque. These results support the theory that unimpaired individuals do not exclusively use central pattern generators or other low-level building blocks to coordinate muscle activity, especially when learning a new task or adapting to external assistance, and demonstrate the potential for using exoskeletons to modulate muscle recruitment and coordination patterns for rehabilitation or performance.     

Application of EMG signals for controlling exoskeleton robots.

Exoskeleton robots are mechanical constructions attached to human body parts, containing actuators for influencing human motion. One important application area for exoskeletons is human motion support, for example, for disabled people, including rehabilitation training, and for force enhancement in healthy subjects. This paper surveys two exoskeleton systems developed in our laboratory. The first system is a lower-extremity exoskeleton with one actuated degree of freedom in the knee joint. This system was designed for motion support in disabled people. The second system is an exoskeleton for a human hand with 16 actuated joints, four for each finger. This hand exoskeleton will be used in rehabilitation training after hand surgeries. The application of EMG signals for motion control is presented. An overview of the design and control methods, and first experimental results for the leg exoskeleton are reported.     

Rehabilitation potential of post-stroke patients training for kinesthetic movement imagination: Motor and cognitive aspects

The rehabilitation potential of post-stroke patients was evaluated after a rehabilitation procedure using a hand exoskeleton controlled via a brain–computer interface (BCI). Examples are given for parameters describing the motor and cognitive functions and the capacity for kinesthetic movement imagination. It is emphasized that instrumental quantitative methods are important to use for adequate assessment of both the rehabilitation potential and the effectiveness of the BCI + exoskeleton procedure.     

Kinematic and dynamic analysis of an anatomically based knee joint

This paper presents a knee-joint model to provide a better understanding on the interaction between natural joints and artificial mechanisms for design and control of rehabilitation exoskeletons. The anatomically based knee model relaxes several commonly made assumptions that approximate a human knee as engineering pin-joint in exoskeleton design. Based on published MRI data, we formulate the kinematics of a knee-joint and compare three mathematical approximations; one model bases on two sequential circles rolling a flat plane; and the other two are mathematically differentiable ellipses-based models with and without sliding at the contact. The ellipses-based model taking sliding contact into accounts shows that the rolling-sliding ratio of a knee-joint is not a constant but has an average value consistent with published measurements. This knee-joint kinematics leads to a physically more accurate contact-point trajectory than methods based on multiple circles or lines, and provides a basis to derive a knee-joint kinetic model upon which the effects of a planar exoskeleton mechanism on the internal joint forces and torque during flexion can be numerically investigated. Two different knee-joint kinetic models (pin-joint approximation and anatomically based model) are compared against a condition with no exoskeleton. The leg and exoskeleton form a closed kinematic chain that has a significant effect on the joint forces in the knee. Human knee is more tolerant than pin-joint in negotiating around a singularity but its internal forces increase with the exoskeleton mass-to-length ratio. An oversimplifying pin-joint approximation cannot capture the finite change in the knee forces due to the singularity effect.     

Biomechanical Evaluation of an Exoskeleton for Rehabilitation of Individuals with Parkinson's Disease

Objective: Prior to implementing any robotic device or new rehabilitation technique into clinical practice, it is essential to examine metrics that reflect the instrument's utility and range of action. For this purpose, this study assesses, based on biomechanical data, the ergonomics and action range of a wrist exoskeleton that may be implemented into clinical practice for the rehabilitation of this joint in patients with Parkinson' disease (PD). Materials and Methods: Eleven individuals with rigidity caused by PD participated in the study. Two distinct tasks were proposed: maximum voluntary contraction and sequential wrist flexion and extension; while electromyography and kinematic data were collected in two stages, with and without the exoskeleton. For statistical analysis, Bootstrap resampling, Kolmogorov-Smirnov, and paired Wilcoxon Mann tests were used. Results and Conclusion: Using the exoskeleton resulted in a 22% increase in muscle activation of the carpal extensors and a 9% decrease in wrist extension. According to the biomechanical parameters evaluated, the exoskeleton guaranteed the amplitude considered functional for the wrist joint, indicating good mechanical adequacy for use in clinical practice.     

Principles of neurorehabilitation based on the brain-computer interface and biologically adequate control of the exoskeleton

The neurophysiological prerequisites for the development and operation of the brain-computer interfaces (BCI) that allow cerebral electrical signals alone to control external technical devices are considered. A BCI based on the discrimination of the EEG patterns related to imagery of extremity movements is described. The possibility of the rehabilitation of patients with motor disorders by means of the BCI based on motor imagery and the exoskeleton controlled by it is discussed.     

Iterative learning control applied to a hybrid rehabilitation exoskeleton system powered by PAM and FES

Currently upper limb exoskeleton rehabilitation robots powered by electric motors used in the hospitals are usually cumbersome, bulky and unmovable. Our developed RUPERT is a low-cost lightweight portable exoskeleton rehabilitation robot that can encourage stroke patients with high stiffness in arm flexor muscles to receive frequent intensive rehabilitation trainings in the community or home, but its joints are unidirectionally actuated by pneumatic artificial muscles (PAMs). RUPERT with one PAM of each joint is not suitable for stroke patients with weak muscles in the flaccid paralysis period. Functional electrical stimulation (FES) uses current with low frequency to activate paralyzed muscles, which can produce muscle torque and compensate the unidirectional drawbacks of RUPERT, so as to realize the two-way motion of its joints for passive reaching trainings. As both the exoskeleton robot driven by PAMs and neuromuscular skeletal system under FES possess the highly nonlinear and time-varying characteristics, which adds control difficulty to the hybrid dynamic system, iterative learning control (ILC) is chosen to control this newly designed hybrid rehabilitation system to realize repetitive task trainings.     

Hybrid adaptive optimal control of anaerobic fluidized bed bioreactor for the de-icing waste treatment

Hybrid adaptive control strategy was developed and tested for the degradation of propylene glycol, a major component in de-icing waste, in an anaerobic fluidized bed bioreactor (AFBR). A linearized model with time-varying parameters was first employed to describe the dynamic behavior of the AFBR using a recursive off-line system identification method. A hybrid adaptive control strategy was then tested using a recursive off-line system identification routine followed by an on-line adaptive optimal control algorithm. The objective of the controller was to achieve the desired set point value of the propionate concentration (stand-alone control output variable) by manipulating the dilution rate (control input variable). To do so, the optimal control law was developed by minimizing a cost function with constraint equations. This novel idea was successfully applied to the underlying system for 200 h. The set point (700 mg HPrl(-1)) was achieved even in the case where the feed concentration suddenly increased by 50% (9000 mg HPrl(-1) to 13500 mg HPrl(-1)).     

Invariant hip moment pattern while walking with a robotic hip exoskeleton

Robotic lower limb exoskeletons hold significant potential for gait assistance and rehabilitation; however, we have a limited understanding of how people adapt to walking with robotic devices. The purpose of this study was to test the hypothesis that people reduce net muscle moments about their joints when robotic assistance is provided. This reduction in muscle moment results in a total joint moment (muscle plus exoskeleton) that is the same as the moment without the robotic assistance despite potential differences in joint angles. To test this hypothesis, eight healthy subjects trained with the robotic hip exoskeleton while walking on a force-measuring treadmill. The exoskeleton provided hip flexion assistance from approximately 33% to 53% of the gait cycle. We calculated the root mean squared difference (RMSD) between the average of data from the last 15 min of the powered condition and the unpowered condition. After completing three 30-min training sessions, the hip exoskeleton provided 27% of the total peak hip flexion moment during gait. Despite this substantial contribution from the exoskeleton, subjects walked with a total hip moment pattern (muscle plus exoskeleton) that was almost identical and more similar to the unpowered condition than the hip angle pattern (hip moment RMSD 0.027, angle RMSD 0.134, p<0.001). The angle and moment RMSD were not different for the knee and ankle joints. These findings support the concept that people adopt walking patterns with similar joint moment patterns despite differences in hip joint angles for a given walking speed.     

Modulation of shoulder muscle and joint function using a powered upper-limb exoskeleton

Robotic-assistive exoskeletons can enable frequent repetitive movements without the presence of a full-time therapist; however, human-machine interaction and the capacity of powered exoskeletons to attenuate shoulder muscle and joint loading is poorly understood. This study aimed to quantify shoulder muscle and joint force during assisted activities of daily living using a powered robotic upper limb exoskeleton (ArmeoPower, Hocoma). Six healthy male subjects performed abduction, flexion, horizontal flexion, reaching and nose touching activities. These tasks were repeated under two conditions: (i) the exoskeleton compensating only for its own weight, and (ii) the exoskeleton providing full upper limb gravity compensation (i.e., weightlessness). Muscle EMG, joint kinematics and joint torques were simultaneously recorded, and shoulder muscle and joint forces calculated using personalized musculoskeletal models of each subject’s upper limb. The exoskeleton reduced peak joint torques, muscle forces and joint loading by up to 74.8% (0.113 Nm/kg), 88.8% (5.8%BW) and 68.4% (75.6%BW), respectively, with the degree of load attenuation strongly task dependent. The peak compressive, anterior and superior glenohumeral joint force during assisted nose touching was 36.4% (24.6%BW), 72.4% (13.1%BW) and 85.0% (17.2%BW) lower than that during unassisted nose touching, respectively. The present study showed that upper limb weight compensation using an assistive exoskeleton may increase glenohumeral joint stability, since deltoid muscle force, which is the primary contributor to superior glenohumeral joint shear, is attenuated; however, prominent exoskeleton interaction moments are required to position and control the upper limb in space, even under full gravity compensation conditions. The modeling framework and results may be useful in planning targeted upper limb robotic rehabilitation tasks.     

Recovery of the motor function of the arm with the aid of a hand exoskeleton controlled by a brain–computer interface in a patient with an extensive brain lesion

The dynamics of motor function recovery in a patient with an extensive brain lesion has been investigated during a course of neurorehabilitation assisted by a hand exoskeleton controlled by a brain–computer interface. Biomechanical analysis of the movements of the paretic arm recorded during the rehabilitation course was used for an unbiased assessment of motor function. Fifteen procedures involving hand exoskeleton control (one procedure per week) yielded the following results: (a) the velocity profile for targeted movements of the paretic hand became nearly bell-shaped; (b) the patient began to extend and abduct the hand, which was flexed and adducted at the beginning of the course; and (c) the patient started supinating the forearm, which was pronated at the beginning of the rehabilitation course. The first result is interpreted as improvement of the general level of control over the paretic hand, and the two other results are interpreted as a decrease in spasticity of the paretic hand.     

Neural adaptive control for vibration suppression in composite fin-tip of aircraft

In this paper, we present a neural adaptive control scheme for active vibration suppression of a composite aircraft fin tip. The mathematical model of a composite aircraft fin tip is derived using the finite element approach. The finite element model is updated experimentally to reflect the natural frequencies and mode shapes very accurately. Piezo-electric actuators and sensors are placed at optimal locations such that the vibration suppression is a maximum. Model-reference direct adaptive neural network control scheme is proposed to force the vibration level within the minimum acceptable limit. In this scheme, Gaussian neural network with linear filters is used to approximate the inverse dynamics of the system and the parameters of the neural controller are estimated using Lyapunov based update law. In order to reduce the computational burden, which is critical for real-time applications, the number of hidden neurons is also estimated in the proposed scheme. The global asymptotic stability of the overall system is ensured using the principles of Lyapunov approach. Simulation studies are carried-out using sinusoidal force functions of varying frequency. Experimental results show that the proposed neural adaptive control scheme is capable of providing significant vibration suppression in the multiple bending modes of interest. The performance of the proposed scheme is better than the H(infinity) control scheme.     

State Classification and Motion Description for the Lower Extremity Exoskeleton SJTU-EX

A lower extremity exoskeleton, SJTU-EX, is proposed, which mainly aims to help soldiers and workers to support a payload in motions. To solve the issues of the exoskeleton-environment and exoskeleton-human interactions, four types of foot contact are proposed based on their different kinematic characteristics. By the combination of single leg states, sixteen exo- skeleton states are presented using a series of meaningful notations from the topological point of view. The generalized function set （GF set） theory is employed to achieve the kinematic characteristics of the end effectors in different states. Moreover six mathematical formulations of the dynamics of the exoskeleton and its interactions with the human wearer are developed for different exoskeleton states. The applicability and potential of the proposed classification method are demonstrated by analyzing common lower limb motions, which are described concisely by a sequence of notations. Finally, a new concept of characteristic state is put forward to uniquely indicate the type of motion.     

Adaptive Control Based Harvesting Strategy for a Predator–Prey Dynamical System

This paper deals with designing a harvesting control strategy for a predator–prey dynamical system, with parametric uncertainties and exogenous disturbances. A feedback control law for the harvesting rate of the predator is formulated such that the population dynamics is asymptotically stabilized at a positive operating point, while maintaining a positive, steady state harvesting rate. The hierarchical block strict feedback structure of the dynamics is exploited in designing a backstepping control law, based on Lyapunov theory. In order to account for unknown parameters, an adaptive control strategy has been proposed in which the control law depends on an adaptive variable which tracks the unknown parameter. Further, a switching component has been incorporated to robustify the control performance against bounded disturbances. Proofs have been provided to show that the proposed adaptive control strategy ensures asymptotic stability of the dynamics at a desired operating point, as well as exact parameter learning in the disturbance-free case and learning with bounded error in the disturbance prone case. The dynamics, with uncertainty in the death rate of the predator, subjected to a bounded disturbance has been simulated with the proposed control strategy.     

Surface EMG pattern recognition for real-time control of a wrist exoskeleton

Background  

Surface electromyography (sEMG) signals have been used in numerous studies for the classification of hand gestures and movements and successfully implemented in the position control of different prosthetic hands for amputees. sEMG could also potentially be used for controlling wearable devices which could assist persons with reduced muscle mass, such as those suffering from sarcopenia. While using sEMG for position control, estimation of the intended torque of the user could also provide sufficient information for an effective force control of the hand prosthesis or assistive device. This paper presents the use of pattern recognition to estimate the torque applied by a human wrist and its real-time implementation to control a novel two degree of freedom wrist exoskeleton prototype (WEP), which was specifically developed for this work.     

Mechanical performance of artificial pneumatic muscles to power an ankle-foot orthosis

We developed a powered ankle-foot orthosis that uses artificial pneumatic muscles to produce active plantar flexor torque. The purpose of this study was to quantify the mechanical performance of the orthosis during human walking. Three subjects walked at a range of speeds wearing ankle-foot orthoses with either one or two artificial muscles working in parallel. The orthosis produced similar total peak plantar flexor torque and network across speeds independent of the number of muscles used. The orthosis generated approximately 57% of the peak ankle plantar flexor torque during stance and performed approximately 70% of the positive plantar flexor work done during normal walking. Artificial muscle bandwidth and force-length properties were the two primary factors limiting torque production. The lack of peak force and work differences between single and double muscle conditions can be explained by force-length properties. Subjects altered their ankle kinematics between conditions resulting in changes in artificial muscle length. In the double muscle condition greater plantar flexion yielded shorter artificial muscles lengths and decreased muscle forces. This finding emphasizes the importance of human testing in the design and development of robotic exoskeleton devices for assisting human movement. The results of this study outline the mechanical performance limitations of an ankle-foot orthosis powered by artificial pneumatic muscles. This orthosis could be valuable for gait rehabilitation and for studies investigating neuromechanical control of human walking.     

Custom sizing of lower limb exoskeleton actuators using gait dynamic modelling of children with cerebral palsy

The use of exoskeletons as an aid for people with musculoskeletal disorder is the subject to an increasing interest in the research community. These devices are expected to meet the specific needs of users, such as children with cerebral palsy (CP) who are considered a significant population in pediatric rehabilitation. Although these exoskeletons should be designed to ease the movement of people with physical shortcoming, their design is generally based on data obtained from healthy adults, which leads to oversized components that are inadequate to the targeted users. Consequently, the objective of this study is to custom-size the lower limb exoskeleton actuators based on dynamic modeling of the human body for children with CP on the basis of hip, knee, and ankle joint kinematics and dynamics of human body during gait. For this purpose, a multibody modeling of the human body of 3 typically developed children (TD) and 3 children with CP is used. The results show significant differences in gait patterns especially in knee and ankle with respectively 0.39 and ?0.33 (Nm/kg) maximum torque differences between TD children and children with CP. This study provides the recommendations to support the design of actuators to normalize the movement of children with CP.     

Online estimation algorithm for a biaxial ankle kinematic model with configuration dependent joint axes

The kinematics of the human ankle is commonly modeled as a biaxial hinge joint model. However, significant variations in axis orientations have been found between different individuals and also between different foot configurations. For ankle rehabilitation robots, information regarding the ankle kinematic parameters can be used to estimate the ankle and subtalar joint displacements. This can in turn be used as auxiliary variables in adaptive control schemes to allow modification of the robot stiffness and damping parameters to reduce the forces applied at stiffer foot configurations. Due to the large variations observed in the ankle kinematic parameters, an online identification algorithm is required to provide estimates of the model parameters. An online parameter estimation routine based on the recursive least-squares (RLS) algorithm was therefore developed in this research. An extension of the conventional biaxial ankle kinematic model, which allows variation in axis orientations with different foot configurations had also been developed and utilized in the estimation algorithm. Simulation results showed that use of the extended model in the online algorithm is effective in capturing the foot orientation of a biaxial ankle model with variable joint axis orientations. Experimental results had also shown that a modified RLS algorithm that penalizes a deviation of model parameters from their nominal values can be used to obtain more realistic parameter estimates while maintaining a level of estimation accuracy comparable to that of the conventional RLS routine.