Multi-Agent Reinforcement Learning For Swarm Robots Formation

• Main Author: Bujang, Christina
• Format: Monograph
• Language: English
• Published: Universiti Sains Malaysia 2021
• Subjects: T Technology
• Online Access: http://eprints.usm.my/54499/1/Multi-Agent%20Reinforcement%20Learning%20For%20Swarm%20Robots%20Formation.pdf; http://eprints.usm.my/54499/

The project discussed the Multi-Agent Reinforcement Learning (MARL) with an idea to the proposed mobile robot which able to follow the line and avoid the obstacle in a given environment. The reinforcement learning algorithm offers one of the most general frameworks in learning subjects to address some of the control issues in a multi-agent system. The mobile robot is an independent agent that can use sensors, actuators, and control techniques to navigate intelligently based on the specific task required. Specifically, reinforcement learning is employed for developing the training process for the mobile robot to reach the given task as it needs to learn by itself to follow the black line and avoid the obstacle in a given environment based on this project proposed. The reinforcement learning approach presents the algorithm for MARL in a cooperative problem to improve control performance. Experimental and simulation will be carried out to validate the results of the multi-agent control performance. Hence, it should be easy to observe if the control performance shows improvement after learning and can achieve the project proposed. The experiment will therefore indicate the results of the simulation and apply it to the real-time environment as proposed by the project.