Path planning for coherent and persistent groups

Abstract

This paper addresses the problem of group path planning while maintaining group coherence and persistence. Group coherence ensures that a group minimizes both longitu- dinal and lateral dispersion, and is achieved with the introduc- tion of a deformation penalty to the cost formulation. When the deformation penalty is significantly high, a group may split and later merge. Group persistence is modeled by introducing split and merge actions in the action space, and adding a split penalty to the cost measure. We formulate the problem domain (state, action space, and cost formulation), present our path planning approach for coherent and persistent groups, and provide empirical results demonstrating the capabilities of our method on a variety of challenging scenarios. I. INTRODUCTION Global navigation for autonomous agents in complex environments is well studied, with many proposed solutions. However, path planning for groups is still an active research with many open problems that are yet to be addressed. Agents within a group share a common target and must try to stay together by satisfying constraints on lateral and longitudinal dispersion, thus maintaining group coherence. Additionally, a group must remain persistent unless the environment demands group splitting and reformation. These dispersion constraints and the ability to split and reform need to be modeled at the global planning level, producing an optimal navigation strategy that minimizes distance traveled, group deformation, and split penalty. This paper presents a path planning approach for coherent and persistent groups in arbitrarily complex environments. The navigable regions in the environment are represented using a triangulated navigation mesh (1) with precomputed local clearance information. A group is represented as a shape constrained area, which incurs a deformation penalty when it deviates from its rest shape, and can split (and reform) in necessary situations. The group action space is extended to include split and merge actions, and a defor- mation cost and a split penalty are introduced into the cost formulation of the search. The cost due to deformation is modeled as the lateral and longitudinal dispersion of the group from its rest shape, while split penalty is computed using the current split status of the group (number of splits