Feedback-Error Learning Control for Powered Assistive Devices

Active orthoses (AOs) are becoming relevant for user-oriented training in gait rehabilitation. This implies efficient responses of AO’s low-level controllers with short time modeling for medical applications. This thesis investigates, in an innovative way, the performance of Feedback-Error Learning (FEL) control to time-effectively adapt the AOs’ responses to user-oriented trajectories and changes in the dynamics due to the interaction with the user. FEL control comprises a feedback PID controller and a neural network feedforward controller to promptly learn the inverse dynamics of two AOs. It was carried out experiments with able-bodied subjects walking on a treadmill and considering external disturbances to user-AO interaction. Results showed that the FEL control effectively tracked the user-oriented trajectory with position errors between 5% to 7%, and with a mean delay lower than 25 ms. Compared to a single PID control, the FEL control decreased by 16.5% and 90.7% the position error and delay, respectively. Moreover, the feedforward controller was able to learn the inverse dynamics of the two AOs and adapt to variations in the user-oriented trajectories, such as speed and angular range, while the feedback controller compensated for random disturbances. FEL demonstrated to be an efficient low-level controller for controlling AOs during gait rehabilitation.

[1]  E H F van Asseldonk,et al.  Speed-dependent reference joint trajectory generation for robotic gait support. , 2014, Journal of biomechanics.

[2]  Mitsuo Kawato,et al.  Feedback-error-learning neural network for trajectory control of a robotic manipulator , 1988, Neural Networks.

[3]  Khairul Anam,et al.  Active Exoskeleton Control Systems: State of the Art , 2012 .

[4]  Sebastian Ruder,et al.  An overview of gradient descent optimization algorithms , 2016, Vestnik komp'iuternykh i informatsionnykh tekhnologii.

[5]  Magdo Bortole.  Robotic exoskeleton with an assist-as-needed control strategy for gait rehabilitation after stroke , 2014 .

[6]  Jose Gonzalez-Vargas,et al.  Feedback Error Learning Controller for Functional Electrical Stimulation Assistance in a Hybrid Robotic System for Reaching Rehabilitation , 2016, European journal of translational myology.

[7]  Xiaodong Liu,et al.  The PID Controller Based on the Artificial Neural Network and the Differential Evolution Algorithm , 2012, J. Comput..

[8]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[9]  Tingfang Yan,et al.  Review of assistive strategies in powered lower-limb orthoses and exoskeletons , 2015, Robotics Auton. Syst..

[10]  Magali R. G. Meireles,et al.  A comprehensive review for industrial applicability of artificial neural networks , 2003, IEEE Trans. Ind. Electron..

[11]  Robert Riener,et al.  Control strategies for active lower extremity prosthetics and orthotics: a review , 2015, Journal of NeuroEngineering and Rehabilitation.

[12]  Wei Meng,et al.  Recent development of mechanisms and control strategies for robot-assisted lower limb rehabilitation , 2015 .

[13]  Slavka Viteckova,et al.  Wearable lower limb robotics: A review , 2013 .

[14]  Mitsuo Kawato,et al.  Feedback-Error-Learning Neural Network for Supervised Motor Learning , 1990 .

[15]  J. Sola,et al.  Importance of input data normalization for the application of neural networks to complex industrial problems , 1997 .

[16]  J. Moreno,et al.  The H2 robotic exoskeleton for gait rehabilitation after stroke: early findings from a clinical study , 2015, Journal of NeuroEngineering and Rehabilitation.

[17]  R. Riener,et al.  Virtual reality to control active participation in a subacute stroke patient during robot-assisted gait training , 2011, 2011 IEEE International Conference on Rehabilitation Robotics.

[18]  N. Hoshimiya,et al.  Joint angle control by FES using a feedback error learning controller , 2005, IEEE Transactions on Neural Systems and Rehabilitation Engineering.