Intelligent active force control of a human-like arm actuated by pneumatic artificial muscles

Abstract

Robotic system driven by fluidic muscles is time-varying and exhibits high degree of nonlinearity due to the internal system behaviours and thus controlling it constitutes a major problem. This poses a great challenge to researchers to come up with suitable technique/s to control such system effectively. The research presents a simulation and experimental study of a force control method applied to a two-link planar ‘human-like’ robot arm that is actuated by fluidic muscles. Active force control (AFC) based scheme was particularly implemented to the system incorporating two types of intelligent techniques, namely, fuzzy logic (FL) and iterative learning (IL) to effectively and robustly control the arm driven by a pneumatic artificial muscle (PAM) system and subject to a number of operating and loading conditions. The PAM is actuated by two groups of fluidic muscles in bicep/tricep configuration. The simulation and the experimental study verify that proposed system performs excellently even in the presence of uncertainties, hysteresis behaviour of the actuator and inherent nonlinearity. Joint trajectory planning was applied to ensure that the arm tracks the given input commands accurately considering a number of different frequency settings. The simulated system was complemented and validated through an experimental study carried out on a developed rig via a convenient hardware-in-the-loop simulation (HILS) technique using suitable hardware and software interface. The obtained results both through simulation and experimental investigation clearly imply the viability of the proposed AFC-based system in controlling the PAM actuated robotic arm and they also demonstrate the system superiority over the PID controller alone counterpart.