Robotic Arm Drift Compensation Using Adaptive Computed Torque Control Based on Reinforcement Learning

Robotic arms often experience drift over time due to various factors such as temperature fluctuations, gearbox backlash, wear and tear, sensor inaccuracies, and changes in load. While model-based methods have addressed drift compensation, they come with considerable limitations compared to data-driven approaches. This paper will review the PID, Computed Torque Control and Adaptive Computed Torque Control based on Reinforcement Learning, soft actor critic (SAC) model, examining their strengths, weaknesses, and potential areas for improvement to enhance the accuracy and precision of robotic systems in industrial settings.