Research on task allocation optimization of multi-UAV logistics collaborative distribution under low-altitude economic background

• 1、School of Management, Wuhan University of Science and Technology, Wuhan 430065

Abstract：With the rise of low-altitude economy as a national strategic industry, urban UAV logistics and distribution is facing the challenge of collaborative scheduling brought by the densification of airspace. The increase in the number of Uavs not only increases the complexity of task allocation, but also leads to the intensification of path conflict problem, which further threatens flight safety and scheduling efficiency. To solve this problem, this study proposes a hybrid intelligent optimization framework (GA-QL) integrating genetic algorithm and Q-learning reinforcement learning. In this framework, the delivery completion time and the number of path conflict points are simultaneously optimized by the dual-objective optimization model under the actual constraints. At the level of algorithm design, the improved PMX crossover operator and random disturbance mutation strategy are used to optimize the task allocation scheme of "distribution center-UAV-customer point". The Q-learning module is innovatively introduced to monitor the convergence state of the algorithm in real time, dynamically adjust the parameters of crossover rate and mutation rate, and effectively avoid the problem of premature convergence. Experimental results show that GA-QL framework has significant advantages in path conflict control and computational efficiency compared with traditional algorithms in various task distribution scenarios, and can reduce the fitness value by up to 8.13%. In addition, the effectiveness of reinforcement learning for genetic algorithm improvement is further verified through ablation experiments. With the expansion of problem scale, the improvement efficiency is more obvious.

Keywords： multi-UAV Assignment of tasks Q-learning Genetic algorithm