Authors: Irving Solis

Venue: Doctoral dissertation
DOI: NA
Link to Publication

Abstract:
Hybrid multi-robot pathfinding algorithms have proven effective in solving discrete problems with numerous robots, using a blend of independent planning and coordinated synchronization to avoid collisions. While these algorithms excel in discrete environments, they encounter significant challenges in continuous multi-robot motion planning due to the complexity of variable positions and motions. This dissertation introduces a set of techniques to extend hybrid planning to continuous scenarios. We first adapt a state-of-the-art discrete pathfinding method to operate on sampling-based roadmaps, improving scalability and solution quality. To address the limitations in complex systems requiring precise coordination, we developed a dynamic approach that adjusts coordination levels based on the scenario's needs. This approach seamlessly transitions between decoupled and coupled planning spaces, focusing computational resources where coordination is most critical. Building on this adaptive strategy, we also present an experience-based method that learns from past conflict resolutions. By storing successful solutions in a compact database, our method reuses these solutions in similar scenarios, reducing the need for replanning. Together, these methods advance hybrid multi-robot motion planning by bridging the gap between discrete and continuous problems, paving the way for more versatile and adaptive solutions to real-world challenges. 

@phdthesis{Solis-hmmp-2024, 
 address = {College Station, TX}, 
 author = {Juan Irving Solis Vidana}, 
 school = {Texas A\&M University}, 
 title = {Hybrid Multi-Robot Motion Planning}, 
 type = {PhD dissertation}, 
 url = {https://parasollab.web.illinois.edu/publications/papers/solis-thesis.pdf}, 
 year = {2024} 
}