Adaptive backstepping control of pneumatic anthropomorphic hand

Abstract

This research presents a nonlinear adaptive backstepping strategy for control of a pneumatic anthropomorphic robotic hand. An anthropomorphic hand with three fingers has been developed in this work. The fingers are driven by tendons and actuated by human muscle-like actuators known as Pneumatic Artificial Muscle (PAM). The high nonlinear dynamics of these actuators and the inherent hysteresis in their behaviour lead to the modelling and control problems that cause a lack of robustness in the hand`s performance. The robotic finger and the PAM actuator have been mathematically modelled as a nonlinear second order system based on an empirical approach. An adaptive backstepping controller has been designed in two design steps based on the nonlinear second order system model for position control of the pneumatic anthropomorphic hand. In the design procedure the estimator of the system uncertainty is incorporated to the proposed control law which is extended for grasping objects with changing weights using a slip detection strategy. In addition, a cascade control system is developed by combining a conventional PID control, as the inner pressure control loop, with the adaptive backstepping control as the outer position control loop. Simulation and experimental test have been conducted using an experiment setup to evaluate the performance of the designed controller. Based on both simulation and experimental results, the adaptive backstepping position controller is capable to compensate the uncertain coulomb friction force of PAM actuator achieving the desired angular trajectory with RMSE of hysteresis behaviour in range of 0.09o - 0.18o and RMSE of angular position control in range of 0.05o - 0.11o. The cascade controller has shown a stable supply of pressurized air with average settling time of 0.38 s - 0.57 s. In terms of force control, the robotic hand is able to maintain grasping objects when their weight is increased up to 500 g by detecting the slip signal generated by the force sensor. Therefore, based on the obtained results, the controller is capable of tracking the desired position accurately and the pneumatic anthropomorphic hand is able to prevent the object from dropping when its weight is increased. For future researches, the adaptive backstepping position controller can be used to overcome other uncertain parameters such as the viscous friction. The further pneumatic hand can also be improved by increasing the number of the robotic fingers and DOFs to improve its manipulation and grasping ability.