Mobile Robot Collision Avoidance Based on Deep Reinforcement Learning
Miss Yuting Tao (M.Phil. candidate)
Department of Mechanical Engineering
The University of Hong Kong
Date & Time
Wednesday, 29 March 2023
Room 7-34, Haking Wong Building, HKU
Mobile robot obstacle avoidance is an important problem in robot navigation and movement. In practical applications, robots need to avoid obstacles and reach the target position to complete various tasks, such as patrol, cleaning, transportation, and so on. Traditional obstacle avoidance methods are usually based on rules or PID controllers, which may perform well in specific scenarios but are difficult to handle complex environments and changing obstacles. Deep reinforcement learning is a reinforcement learning method based on neural networks that learns optimal strategies by interacting with the environment. In mobile robot obstacle avoidance, deep reinforcement learning can guide robots to learn efficient and flexible obstacle avoidance behaviors by using reward functions. In recent years, the development of deep reinforcement learning algorithms, such as DQN, DDPG, and PPO, has enabled better solutions to robot obstacle avoidance problems. In this seminar, we propose an obstacle avoidance method based on deep reinforcement learning for mobile robots that utilizes dimension-reduced depth images as observations to achieve efficient and reliable collision avoidance in complex environments.