An intelligent force control strategy for soft robotic grippers

Abstract

Abstract: Soft robotics and artificial intelligence are becoming increasingly practical tools to overcome the challenges involved in the manipulation of arbitrary objects but much of the current research focuses on pose estimation and finger placement, with limited research into effectively predicting the force required for safe manipulation. Moreover, there is a lack of closed-loop force control strategies for soft robots. This paper will present a strategy to estimate the required force and position setpoints for a set of objects using machine learning and computer vision techniques. To validate the efficacy of the setpoint predictions, an electrically driven parallel jaw mechanism was used, as well as a position and force controller with a cascaded predictive controller and a mechanism to switch between position and force control. The setpoint estimation strategy was found to have acceptable performance, and the control algorithms showed good control performance.