Research on Obstacle Avoidance and Path of an Intelligent Robot Based on Reinforcement Learning

Abstract

This paper focuses on the application and verification of the Q-learning algorithm and the Sarsa algorithm in a robot obstacle avoidance scenario. With the increasing application of intelligent robots, the complex dynamic environment puts forward higher requirements for their obstacle avoidance ability. Traditional obstacle avoidance algorithms are difficult to adapt to a changing environment. Reinforcement learning shows strong adaptability and obstacle avoidance effects through the interaction between robots and the environment, which has become a current research hotspot. In this paper, based on Q-learning and the Sarsa algorithm, a Python program is used to build the experimental environment, and the test scene is processed graphically to facilitate the observation of the obstacle avoidance path of the robot. Both Q-learning and the State-Action-Reward-State-Action (SARSA) algorithm avoid conventional obstacles and reach the end point by the shortest path in the experiment. In the dangerous obstacle scene, the Q-learning algorithm can still avoid obstacles and find the shortest path, while the Sarsa algorithm selects a longer route. The verification results show that the two algorithms have their advantages and disadvantages, which provides a reference for the selection and optimization of robot obstacle avoidance algorithms and has important practical significance and theoretical value. This study aims to promote the development of robot obstacle avoidance technology and provide a useful reference for research and application in related fields.