Fast algorithms for UAV tasking and routing

Abstract

The problem of partitioning a set of geographically distributed tasks among a team of UAVs that must travel to perform the tasks arises in many contexts, from surveillance to environmental monitoring. When UAVs have limits in their travel distances, it is important to select subsets of tasks for each UAV that achieve highest team values while remaining feasible given limits on UAV resources. This class of problems has been studied previously in computer science and operations research as orienteering problems. This problem is known to be NP-Hard. Previous algorithms for solving orienteering problems as integer programming problems are slow; faster algorithms have been proposed based on heuristic techniques used in traveling salesperson problems. In this paper, we develop a different class of fast algorithms, based on a decomposition of the orienteering problem into a knapsack assignment problem and a subsequent traveling salesperson problem. We combine greedy algorithms for knapsack problems with the use of spanning trees to estimate traveling salesperson tour lengths to obtain new, approximate algorithms for orienteering problems. We conduct experiments with single and multiple UAV algorithms on benchmark problems as well as problems derived from a LANDAT dataset, comparing our new algorithms with other algorithms in the literature. Our results indicate that our new algorithms are consistently fast, and perform better than competing algorithms.