The Method for Automatic Adjustment of AGV’s PID Based on Deep Reinforcement Learning

Abstract

This paper proposes a method for automatic adjustment of PID(Proportion Integration Differentiation) based on deep reinforcement learning in order to solve the problem of smooth movement of AGV(Automated Guided Vehicle). First, based on reinforcement learning, the problem of AGV smooth operation is transformed into the solution of PID adjustment operation sequence. The action value model is constructed using the Deep Q-Learning Network(DQN) algorithm. Then, take the AGV adjustment PID as the research object. The goal is to achieve smooth movement of AGV. The specific design of AGV automatic PID adjustment method based on deep reinforcement learning is introduced. Finally, the simulation data is used to train the AGV action value network model. The method is validated in the ROS(Robot Operating System) simulation environment. Then compare with the actual environment. The results show that the automatic adjustment method of AGV’s PID based on deep reinforcement learning can adjust PID to the optimum without manual work. It can make AGV run smoothly. The feasibility and effectiveness of the method are illustrated and the problem of automatic PID determination in AGV movement control is solved.

An overview of the development of key technologies in autonomous driving technology

“…Meiling Chen presents a deep reinforcement learning method based on a self-attention mechanism, aiming at the problems of weak feature extraction ability and low training efficiency when a reinforcement learning model is used to extract high-dimensional data [7]. A deep reinforcement learning method based on an information bottleneck is proposed to solve the problem that the reinforcement learning algorithm only completes the policy update according to the reward size while ignoring the influence of the feature extraction ability of the model on the strategy learning.…”

Autonomous Navigation of Robots: Optimization with DQN

“…The robot can be trained by interacting with its environment and receiving feedback in the form of rewards or penalties based on the actions it takes. For example, the robot may receive a reward for successfully navigating to a particular location, while receiving a penalty for colliding with an obstacle [33][34][35][36][37]. Over time, the robot can use the feedback it receives to learn the optimal path to take in different environments.…”