Authors: Hannah Lee, James Motes, Zachary Serlin, Marco Morales, Nancy M. Amato

Venue: Extended Abstract for IEEE International Conference on Robotics and Automation @ 40 (ICRA@40)
DOI:
Link to Publication

Find PDF here: PDF

Abstract:
Decentralized multi-agent pathfinding (MAPF) is a crucial area of study within artificial intelligence and robotics, focusing on enabling multiple autonomous agents to navigate a shared environment without conflicts. This approach is valuable for applications such as warehouse automation and vehicle routing. Decentralized MAPF distributes the decision-making process among individual agents, enhancing scalability, robustness, and flexibility. It is particularly important for managing large-scale dynamic environments where centralized control is impractical or intractable. Decentralized MAPF faces significant challenges under multi-hop communication assumptions, such as complexities in ensuring timely and accurate data exchange. The assumption that networks can manage planning without assigning specific communication routes or planning roles often fails in practice. To address these shortcomings, we leverage concepts from Hierarchical Composition Conflict-Based Search (HC-CBS) to introduce our extension, Hierarchical Composition for Multi-hop Distributed Communication (HCMDC). This approach strategically determines communication paths within distributed networks along which the MAPF problem is built and solved. By doing so, we consider the intricacies of multi-hop communication while ensuring safe and efficient navigation. Our preliminary experiments show the advantages of distributed planning in random environments.

```

Categories:

Updated: