Interactive approaches for bi-objective UAV route planning in continuous space

Türeci, Hannan
We study the route planning problem of unmanned air vehicles (UAVs). We consider two objectives; minimizing total distance traveled and minimizing total radar detection threat since these objectives cover most of the other related factors. We consider routing in a two-dimensional continuous terrain, in which we have infinitely many efficient trajectories between target pairs. We develop interactive algorithms that find the most preferred solution of a route planner (RP), who has either of the underlying preference function structures: linear or quasiconvex. To implement the algorithms to route planning problems, we use approximated nondominated frontiers of the trajectories between targets. In the linear case, we search for supported efficient solutions in two stages. In the first stage, we find the best trajectory between each target pair. In the second stage, we find the tour visiting all targets (traveling salesperson problem, TSP). In the quasiconvex case, we search for both supported and unsupported efficient solutions. We first reduce the objective space to rectangular regions around at most three supported efficient solutions. We then search inside these rectangular regions to find supported/unsupported efficient solutions and narrow our search region. We proceed with pairwise comparisons from the RP and reduce our search space until the two solutions to be compared are close enough. To generate random problem instances, we develop a mathematical model that randomly locates radars in a terrain with known target locations. We then demonstrate the interactive algorithm developed for linear preference functions on two randomly generated problems.