AntSearcher: An Ant Colony Optimization-Based Framework for Moving Target Search with UAVs

Qingxia Li, Ani Dong, Yuhui Zhang, Wenhong Wei

Keywords: Ant Colony Optimization; UAV; Moving Target Search; Swarm Intelligence

TL;DR: AntSearcher is a tensor-based Ant Colony System that efficiently solves the moving-target search problem by combining a dynamic "Telescope" heuristic and adaptive pheromone reinitialization.

Abstract: The Moving Target Search (MTS) problem presents a complex optimization challenge in dynamic environments. This paper introduces AntSearcher, an Ant Colony System (ACS) algorithm tailored to efficiently address the MTS problem. AntSearcher utilizes a custom solution construction graph, where ants traverse edges to identify UAV paths maximizing the probability of detecting moving targets. The algorithm incorporates problem-specific heuristic and pheromone tensors to steer the search process. It includes a time-varying telescope heuristic, which exploits the target's initial belief map to enhance the early search phase, and a reinitialization mechanism to mitigate premature convergence. Extensive experiments across diverse search scenarios demonstrate that AntSearcher surpasses state-of-the-art evolutionary optimizers, showcasing its superior effectiveness and robustness.

Submission Number: 66