Loading...
Multi-agent reinforcement learning approach for robot teaming in Mars exploration
Moon, Sejun
Moon, Sejun
Citations
Altmetric:
Files
Authors
Other Names
Location
Time Period
Advisors
Original Date
Digitization Date
Issue Date
2025-02-26
Type
Abstract
Genre
Keywords
Subjects (LCSH)
Citation
Abstract
Mars exploration is crucial for understanding the planet's potential to support life, its climate and geological history, and for preparing future human missions. However, Mars exploration poses challenges due to constrained communication capabilities. This research proposed a MARL-based method for autonomous path planning for robot-robot teaming. Group-oriented Multi-Agent Reinforcement Learning (GoMARL) is commonly employed for robot-robot teaming. However, this method faces limitations regarding computational complexity as the number of agents increases. The proposed approach designs a hierarchical structure to mitigate this complexity and enhance scalability. The proposed MARL structure organizes agents into multiple layers, where higher-level agents manage strategic decisions like group formation while lower-level agents focus on tactical execution like navigation, thereby reducing computational complexity and improving scalability for large teams of robots. Research ongoing. The experimental environment is set up in ROS2 to simulate Martian conditions. The preliminary reinforcement learning model showed promising results. The proposed method allows multiple robots to navigate autonomously with limited human interaction under extreme circumstances, like Mars. Kansas is highly prone to disasters like Tornados, and the proposed method can contribute to post-disaster rescue.
Table of Contents
Description
Poster project completed at College of Engineering, School of Computing.
Presented at the Kansas Undergraduate Student Research Day at the Capitol, Topeka, KS, February 26, 2025.
Sponsored by Undergraduate Research and Creative Activity Hub, Dorothy and Bill Cohen Honors College.
Publisher
Wichita State University