Integrating vehicle routing and motion planning

Abstract

There has been much interest recently in problems that combine high-level task planning with low-level motion planning. In this paper, we present a problem of this kind that arises in multi-vehicle mission planning. It tightly integrates task allocation and scheduling, who will do what when, with path planning, how each task will actually be performed. It extends classical vehicle routing in that the cost of executing a set of high-level tasks can vary significantly in time and cost according to the low-level paths selected. It extends classical motion planning in that each path must minimize cost while also respecting temporal constraints, including those imposed by the agent's other tasks and the tasks assigned to other agents. Furthermore, the problem is a subtask within an interactive system and therefore must operate within severe time constraints. We present an approach to the problem based on a combination of tabu search, linear programming, and heuristic search. We evaluate our planner on representative problem instances and find that its performance meets the demanding requirements of our application. These results demonstrate how integrating multiple diverse techniques can successfully solve challenging real-world planning problems that are beyond the reach of any single method. Copyright © 2012, Association for the Advancement of Artificial Intelligence. All rights reserved.