A Trajectory Tracking Control Method of Baxter Manipulator Based on Reinforcement Learning

Abstract

A trajectory tracking control method for the Baxter manipulator based on reinforcement learning. First, system identification is performed on the first three joints of the Baxter manipulator, and its continuous-time state-space equation is determined and discretized to obtain a discrete state-space model. This step It is only used to obtain the position and velocity tracking errors of the first three joints of the robot at the next moment during simulation; first, given an initial state of the first three joints of the manipulator, measure and record the next moment of the three joints according to a fixed sampling time Position and speed tracking error, after preprocessing the collected position and speed information, use the recursive least squares method to calculate the weight matrix H corresponding to the optimal control strategy, and finally calculate the optimal feedback at the next moment according to the weight matrix control. The invention automatically adapts to model errors caused by model changes and improves the accuracy of the robot in daily use.

Description

technical field [0001] The invention belongs to the field of intelligent control of manipulators, and specifically provides a Baxter manipulator trajectory tracking control method based on reinforcement learning, which can calculate the optimal control method through the reinforcement learning strategy iteration method when the manipulator model is unknown. A control strategy to reduce the trajectory tracking error, thereby minimizing the loss function of the robotic system. Background technique [0002] In recent years, reinforcement learning theory has received extensive attention and research in the field of robot control. As a common tool in industrial production, industrial robotic arms are widely used in automatic production lines. How to apply the reinforcement learning theory to the motion control of the industrial manipulator, so that it has a certain ability of self-learning, is of great significance to expand the application of the manipulator and reduce the diff...

AI Technical Summary

Problems solved by technology

However, the traditional technology has limitations, which are caused by the uncertainty of the actual system, including the uncertainty of the system model and the disturbance outside the system

When using traditional methods, a model of the system is required, and the accuracy of the model directly affects the accuracy of the control. Even if the model is available, the state feedback controller obtained based on the model is only suitable for an approximate model of the real system dynamics

In addition, the optimal control of the time-varying system is difficult to operate in the actual system, the cost is high, the performance is average, and the actual use value is low. Therefore, through the data-driven method, the input and output data of the system are used to calculate the optimal control of the system. Optimal control is clearly necessary

Images