Iterative Motion Planning in Multi-agent Systems with Opportunistic Communication under Disturbance
Neelanga Thelasingha, Agung Julius, James Humann, James Dotterweich
TL;DR
This work targets uncertainty and asymmetric knowledge in multi-agent motion planning under opportunistic communication. It develops a rigorous mathematical framework using a multi-agent transition system $S=(X,T)$ and a Task Site Assignment (TSA) planner, introducing plan and eigen trajectories along with projection operators to model information delays. A key contribution is the n-step recoverability concept, providing conditions under which disturbances can be corrected and task satisfaction guaranteed, even with asynchronous replanning and limited communications. The authors also propose planning algorithms that enforce synchronization constraints and handle disturbances, yielding bounded performance degradation. Experimental validation on a UAV–UGV coordination task demonstrates the feasibility and effectiveness of the approach in realistic mobility and energy-constrained settings.
Abstract
In complex multi-agent systems involving heterogeneous teams, uncertainty arises from numerous sources like environmental disturbances, model inaccuracies, and changing tasks. This causes planned trajectories to become infeasible, requiring replanning. Further, different communication architectures used in multi-agent systems give rise to asymmetric knowledge of planned trajectories across the agents. In such systems, replanning must be done in a communication-aware fashion. This paper establishes the conditions for synchronization and feasibility in epistemic planning scenarios introduced by opportunistic communication architectures. We also establish conditions on task satisfaction based on quantified recoverability of disturbances in an iterative planning scheme. We further validate these theoretical results experimentally in a UAV--UGV task assignment problem.