Robust control of a system with a pneumatic spring

Abstract

Recently, series elasticity has been realized using pneumatics in human-robot interaction systems. Pneumatic circuits provide not only a flexible power transmission, but also the elastic element in a series elastic actuator (SEA). Pneumatic series elastic systems involve more than twice the number of parameters that influence system behaviors in comparison with rigid robotic systems. In this study, a position controller that eliminates the need of identifying a system model by employing the time delay estimation (TDE) technique is proposed for pneumatic SEA systems. The TDE technique is effective in compensating for system dynamics and all uncertainties involved in system behaviors without imposing computation load. TDE error is cancelled out through a learning way, which improves control performance and leads to asymptotic stability. A simulation study demonstrates the robustness of the proposed controllers against uncertainties imposed on the motor system as well as uncertainties on the end-effector. The simulation shows the efficacy of the learning compensation for TDE error.