Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

RAILGUN: A Unified Convolutional Policy for Multi-Agent Path Finding Across Different Environments and Tasks

Created by
  • Haebom

Author

Yimin Tang, Xiao Xiong, Jingyi Xi, Jiaoyang Li, Erdem B{\i}y{\i}k, Sven Koenig

Outline

This paper presents RAILGUN, the first centralized learning-based policy for the multi-agent pathfinding (MAPF) problem. Unlike existing distributed learning-based methods, RAILGUN utilizes a CNN-based architecture to design a supervised policy, enabling generalization across a wide range of map sizes and agent counts. The model is trained using supervised learning using trajectory data collected from rule-based methods. Extensive experimental results demonstrate that RAILGUN outperforms existing methods and achieves excellent zero-shot generalization performance.

Takeaways, Limitations

•
Takeaways:
◦
We present the first centralized learning-based policy for the multi-agent pathfinding problem.
◦
Demonstrating generalizability across different map sizes and agent counts with map-based policies.
◦
Experimentally verified that zero-shot generalization performance is excellent.
◦
Achieve superior performance compared to existing methods
•
Limitations:
◦
Rely on supervised learning approaches that use data collected from rule-based methods
◦
Lack of detailed description of RAILGUN's architecture and training process (additional information needed)
◦
Lack of validation of generalization performance for other types of MAPF problems (e.g., dynamic environments).
👍
Made with Slashpage