Asymmetric 2-Opt Scheduling for Roadmap-Based Task Planning in Urban Terrain

Abstract

This work presents an automated task scheduling approach for mission planning that allows for task-level motion planning with an unmanned rotorcraft in complex terrain. Given a set of mission task elements (e.g. single waypoints or areas of interest) a 2-opt-based heuristic optimization is chosen to reduce tour costs and achieve 2-optimal results. The 2-opt approach is modified to account for asymmetric partial tour costs derived from a sampling-based path planner. To aid this local optimization a global meta-heuristic guidance like simulated-annealing is evaluated. The iterative convergence is evaluated for different configurations of tour initialization methods, a meta heuristic and two 2-opt variants. Although in theory local minima can exist, given the planning uncertainties and an expected number of tasks of up to 100, the cost sub-optimality of generated task schedules are practically negligible. Furthermore, we emphasize the combination of task scheduling with a multi-query planner in fair comparison against a single-query variant of our approach. For tasks distributed across complex terrain, the combination of 2-opt and our multi-query planner results in feasible tours within the order of a few seconds. Moreover, due to a local optimization type of tour improvement, an extension for online re-scheduling is possible.