A hybrid UCB-SA algorithm for multi-UAV path planning in plant protection

• Clustering and greedy algorithms boost initial solution quality. • Introducing the Upper Confidence Bound algorithm improves the algorithm’s performance. • Experiments confirm UCB-SA excels in plant protection path planning. Path planning for multiple dispersed farmlands is a key challenge in smart agriculture. Existing methods for agricultural machinery often suffer from suboptimal solutions, poor stability, and inefficient scheduling, which limit operational efficiency and resource utilization. To address these issues, this study establishes a mixed-integer linear programming (MILP) model for the collaborative path planning of multiple plant protection unmanned aerial vehicles (UAVs) and proposes a hybrid optimization algorithm that integrates the upper confidence bound (UCB) strategy with simulated annealing (SA). The algorithm improves initial solutions through a clustering-based approach combined with a greedy heuristic, and further enhances search efficiency via a UCB-based perturbation mechanism, accelerating convergence and improving stability. Simulation results show that the proposed algorithm outperforms traditional methods, achieving average cost reductions of 8 %, 13 %, and 23 % compared to standard SA, large neighborhood search (LNS), and genetic algorithm (GA), respectively.