Distributed Model-Based Diffusion: Finite Horizon Contraction under Bounded Delay
Robotics
Summary
The authors study how to plan coordinated paths for many agents, like drones or self-driving cars, which is usually very hard due to complex math and delays in communication. They analyze a technique called Distributed Model-Based Diffusion, which is a way for agents to plan their moves while handling tricky problems and communication lag. They prove this method stays stable and works well even with delays and test it in driving and aerial combat tasks. Their results show improvements over a centralized approach despite the added latency. This means their method could better handle real-world multi-agent scenarios.
Authors
Seth Golembeski, Keith L. Gibson, Alexander Gross, Shreyas Kousik, Anirban Mazumdar
Abstract
Simultaneously optimizing the trajectories of multiple agents is a challenging problem plagued by nonlinearity, nonconvexity, and the curse of dimensionality. A collection of interacting aerial vehicles or self-driving cars in an intersection are examples of complex multi-agent systems that remain difficult to solve without many simplifying assumptions. The presence of communication latency between agents further increases the difficulty. In this paper, we analyze Distributed Model-Based Diffusion: a sampling-based Model-Predictive Control method suitable for highly nonlinear, nonconvex, nonsmooth, multi-agent systems. We prove contraction and robustness to latency for multi-agent, nonconvex problems, showing applicability to real-world constraints. We test the algorithm on a circleswap task, a cooperative medium-fidelity driving task, and in an aerial combat scenario. Despite the addition of latency, our algorithm improves circleswap makespan by 31% and increases aerial combat win rate by 25% compared to centralized Model-Based Diffusion.