Research on Trajectory and Path Planning of Mobile Manipulator Based on Reinforcement Learning

Abstract

Traditional robot trajectory planning algorithms rely on kinematic models and are unable to adapt to the production requirements of dynamic changes in the environment. However, reinforcement learning algorithms do not need to build complex mathematical models and directly train agents to interact with the environment through data, which is highly flexible and more adaptable to the environment. The focus of the art is to design an effective control law for the mobile manipulator, which can be used to perform manipulation tasks such as picking up objects or moving objects in space. The main idea behind this method is to use a model free method called reinforcement learning (RL) to design the optimal control law of the robot, so that it can learn how to move according to the desired trajectory without knowing the dynamics of its environment in advance.