Reinforcement Learning Control Of A Shape Memory Alloy-Based Bionic Robotic Hand

Shape Memory Alloy (SMA) with the advantage of high power-to-weight ratio as a promising alternative to traditional actuators is widely applied to robotic systems. Due to the merits of low noise, driving voltages and small size SMA is used in robotic hand. Hysteresis phenomenon of SMA and dynamic model of the SMA-based robotic hand are complexities in control system. In this paper, we apply a reinforcement learning (RL) algorithm to the robotic hand actuated by SMA for motion control. The proposed hand can achieve the desired bending state and grasp the object with desired bending state effectively and steadily.