Amine M. Mamou and Nadia Saadia
Robotic rehabilitation, PID control, feed-forward neural network control, lowers limbs, human physiological movements, kinematic model, path tracking
Different lower limbs robotic rehabilitation devices are under study to assist therapists in their work. They are particularly useful to recover gestures altered, affected or lost after traffic accident injury, illness or disability. In this paper, we present our contribution to the robotic therapy improvement. We present a developed robotic rehabilitation device dedicated to the lower limbs rehabilitation. Using functional orthotics it will be applied for functional rehabilitation, allowing the reproduction of the physiological joint trajectories and resume segmental loads of bodily movement, especially walking. We have developed a control strategy for this prototype. The aim was to reproduce all types of exercises usually performed through a therapist. We used a decoupled structure where each joint has its own controller. We proposed a new control law using the kinematic model based on a classic proportional-integral-derivative (PID) controller. Although the device implementations have given good results, the outputs contain a lot of fluctuation, which is detrimental to our equipment. For this purpose, we used a feed- forward neural network (FFNN) controller copied onto the first and adding the error derivative as a second input. The drawbacks noted with the PID controller were eliminated with the FFNN controller.
Important Links:
Go Back
