Target-bundled genetic algorithm for multi-unmanned aerial vehicle cooperative task assignment considering precedence constraints

This paper presents a modified genetic algorithm using target-bundle-based encoding and tailored genetic operators to effectively tackle cooperative multiple task assignment problems of heterogeneous unmanned aerial vehicles. In the cooperative multiple task assignment problem, multiple tasks including reconnaissance, attack, and verification have to be sequentially performed on each target (e.g. ground control stations, tanks, etc.) by one or multiple unmanned aerial vehicles. Due to the precedence constraints of different tasks, a singular task-execution order may cause deadlock situations, i.e. one or multiple unmanned aerial vehicles being trapped in infinite waiting loops. To address this problem, a target-bundled genetic algorithm is proposed. As a key element of target-bundled genetic algorithm, target-bundle-based encoding is derived to fix multiple tasks on each target as a target-bundle. And individuals are generated by fixing the task-execution order on each target-bundle subject to task precedence constraints. During the evolution process, bundle-exchange crossover and multi-type mutation operators are customized to generate deadlock-free offspring. Besides, the time coordination method is developed to ensure that task-execution time satisfies task precedence constraints. The comparison results on numerical simulations demonstrate that target-bundled genetic algorithm outperforms particle swarm optimization and random search methods in terms of optimality and efficiency.