Functional electrical stimulation for walking: rule based controller using accelerometers

Functional electrical stimulation (FES) can restore walking in paralyzed patients. Rule based control (RBC) is a promising approach for the control of complex musculoskeletal systems using FES. In this paper, we present a method for the design of a RBC for real time control of walking. The controller uses accelerometer data as inputs while the outputs are estimated muscle activations (50 ms ahead in time). It is designed in two steps: 1) the input-output data for machine learning (ML) are generated using biomechanical gait simulations; 2) the rules are determined by applying ML based on the adaptive neuro-fuzzy inference system. The controller is trained and evaluated using the data recorded from an able bodied subject walking at two gait speeds. Results showed that the estimation of muscle activations was satisfactory at the gait speed for which the controller was trained. Moreover, the RBC demonstrated the ability to generalize to the gait speed that was higher/lower then the one actually used for the training.