Adaptive neuro-fuzzy inference system based active force control with iterative learning for trajectory tracking of a biped robot

We investigate the integration of the active force control (AFC) scheme and the adaptive neuro-fuzzy inference system (ANFIS) as an intelligent controller algorithm to address trajectory-tracking problems in robotic systems, The AFC-ANFIS model exploits iterative learning (IL) to improve the tracking performance based on its trained model. The ANFIS parameters are tuned using both particle swarm optimisation (PSO) and beetle antennae search (BAS) algorithms. The simulation results of two different robots, i.e. a five-link biped robot and a PUMA 560 robot arm, indicate that the proposed AFC-ANFIS controller performs well for trajectory tracking and disturbances rejection. The AFC-ANFIS performance is evaluated and compared with those from other controllers using the average tracking error (ATE) metric. The comparative results reveal that AFC-ANFIS offers a viable approach with a rapid training process to undertaking trajectory tracking and disturbance rejection tasks.