Dynamic Assignment in Distributed Motion Planning with Limited Information

Abstract

Distributed motion planning of multiple agents raises fundamental and novel problems in control theory and robotics. In this paper, we consider the problem of designing distributed motion algorithms that dynamically assign targets or destinations to multiple homogeneous agents. We achieve this goal using a novel control decomposition. In particular, navigation of every agent to any available destination is due to distributed multi-destination potential fields, while the mutual exclusion property of the final assignment is guaranteed by local coordination protocols among the agents. Integration of the proposed controllers results in a hybrid model for every agent, while the overall system is shown to always converge to a valid assignment and have at most polynomial complexity, dramatically reducing the combinatorial nature of purely discrete assignment problems. We conclude by illustrating our approach through nontrivial computer simulations.