Learning-based local search algorithm for load-balanced communication resource allocation in unmanned swarm systems

Unmanned swarm systems (USS) are emerging as a critical architecture for future communication and coordination tasks in dynamic and large-scale environments. This paper investigates the communication resource allocation problem in USS, aiming to minimize the total communication time and balance the load across nodes, while satisfying constraints such as communication range and resource capacity. We formulate this problem as a mixed-integer programming (MIP) model and propose a learning-based local search (LBLS) algorithm that integrates reinforcement learning with tabu search to effectively solve large-scale instances. Experimental results on a set of 34 benchmark instances show that LBLS significantly outperforms the popular iterated local search (ILS) and genetic algorithm (GA). In addition, a convergence analysis is conducted to further verify the robustness of the proposed method.