39th Annual Student Research Conference
Multi-Agent Reinforcement Learning Approach for Robot Teaming in Mars Exploration
Kevin Wang
Prof. Fujian Yan, Faculty Mentor
Mars exploration demands autonomous robotic systems due to communication delays of up to 20 minutes between Earth and Mars, making real-time human control infeasible. This research proposes a hierarchical Multi-Agent Reinforcement Learning (MARL) framework to enable scalable, autonomous robot teaming. While existing Group-oriented MARL (GoMARL) approaches support cooperation, they struggle with scalability and computational complexity as agent counts grow. The proposed hierarchical design addresses this by assigning higher-level agents to team organization and task allocation, while lower-level agents handle local navigation and coordination, reducing redundant communication and enabling larger-scale deployment. The framework will be validated in a PyBullet simulation using Leo Rover models on Mars-like terrain. Results are expected to demonstrate improved adaptability and cooperative exploration. Applications extend beyond Mars to Earth-based scenarios such as disaster response and remote search missions.
Keywords: Multi-Agent Reinforcement Learning (MARL), Autonomous Robot Teaming, Hierarchical Decision-Making, Mars Exploration Simulation, PyBullet Simulation
Topic(s):Computer Science
Presentation Type: Oral Presentation
Session: -3
Location: MG 2007
Time: 10:30