Owing to low cost, high flexibility and delivery efficiency, effectively addressing the challenges of “last-mile” delivery. While collaborative truck-drone delivery systems have been proposed to overcome limitations such as limited battery life and payload capacity, they are not well-suited for large and heavy parcel delivery. To solve the issue, a pioneering heterogeneous multi-drone delivery system. In this system, the mother drone handles the delivery of large and heavy parcels, releasing small drones to manage the delivery of smaller and lighter parcels. To address the complexities of this multi-drone delivery system, we introduce a divide-and-conquer framework consisting of two integral phases. The first phase, the task allocation phase, generates multiple task allocation schemes, while the second phase, the single-drone route planning phase, produces high-quality routes for each individual drone. Two phases are performed in an iterative manner until the predefined stopping criteria are satisfied. In the task allocation phase, we propose a simulated annealing algorithm (SA) to facilitate task allocation among multiple drones, utilizing transfer and recombination operators to generate high-quality solutions. After obtaining the task allocation scheme, a satisfactory route of a mother drone is generated by a variable neighborhood descent algorithm (VND). A desirable route for each single small drone is produced by dynamic programming (DP).Extensive experiments are conducted, demonstrating the outstanding optimization and time efficiency of the proposed two-phase optimization method by the fact that it obtains within a 4.89% gap from the optimal solution generated by CPLEX in 15.48 s for instance up to 125 nodes.

成本低、灵活性高、配送效率高，有效解决“最后一公里”配送的挑战。虽然卡车-无人机协作投递系统已被提出来克服电池寿命和有效负载能力有限等限制，但它们不太适合大型和重型包裹投递。为了解决这个问题，开创性的异构多无人机交付系统。在该系统中，母无人机负责处理大而重的包裹的递送，释放小型无人机来管理更小且轻的包裹的递送。为了解决这种多无人机交付系统的复杂性，我们引入了一个由两个完整阶段组成的分而治之的框架。第一阶段，即任务分配阶段，生成多个任务分配方案，而第二阶段，即单无人机路线规划阶段，为每架无人机生成高质量的路线。以迭代方式执行两个阶段，直到满足预定义的停止标准。在任务分配阶段，我们提出了一种模拟退火算法（SA）来促进多架无人机之间的任务分配，利用转移和重组算子生成高质量的解决方案。获得任务分配方案后，通过可变邻域下降算法（VND）生成母机满意的路线。通过动态规划（DP）为每架小型无人机生成一条理想的路线。进行了大量的实验，证明了所提出的两阶段优化方法具有出色的优化和时间效率，与目标的差距在 4.89% 之内。 CPLEX 在 15.48 秒内生成最佳解决方案，例如最多 125 个节点。