LTL Planning for Groups of Robots

摘要

We approach the general problem of planning and controlling groups of robots from logical and temporal specifications over regions of interest in 2D or 3D environments. The focus of this paper is on planning, and, enabled by our previous results, we assume that the environment is partitioned and described in the form of a graph whose nodes label the partition regions and whose edges capture adjacency relations among these regions. We also assume that the robots can synchronize when penetrating from a region to another. We develop a fully automated framework for generation of robot plans from robot abstract task specifications given in terms of linear temporal logic (LTL) formulas over regions of interest. Inter-robot collision avoidance is guaranteed, and the assignment of plans to specific robots is automatic. The main tools underlying our framework are model checking and bisimilarity equivalence relations