Robust Adaptive PID Control based on Reinforcement Learning for MIMO Nonlinear Six-joint Manipulator

The problem of designing a robust and adaptive PID controller for MIMO nonlinear robotic systems with tiny static error and overshooting has been a long-lasting open challenge. This paper presents a novel adaptive reinforcement learning-based control design for a robot with six joints based on a simple structured PID-like control. The method of proposed hybrid control uses an actor-critic structure and automatically estimates the multiple parameters of the PID controllers. The proposed scheme has been verified in a simulation environment. A comparative experiment demonstrated the successful performance of the method against other adaptive methods for MIMO has indicated.