Multi-Objective Path Optimization for Truck–Drone Collaborative Delivery Based on Subspace Contribution-Driven Adaptive Resource Allocation
This paper proposes RC-NSGA-II, a novel multi-objective optimization algorithm that integrates K-means-based subspace contribution-driven resource allocation, random-key encoding, and adaptive local search to effectively solve the complex three-objective truck-drone collaborative delivery problem, demonstrating superior performance in solution quality and efficiency across various benchmark instances.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the bustling arteries of modern cities, the final leg of a delivery journey—the "last mile"—is often the most expensive and complicated part of the entire logistics chain. It is the stretch where a package moves from a local warehouse to a doorstep, a phase that can consume nearly half of all shipping costs. For decades, this task has relied almost entirely on trucks, which are reliable but often slow to navigate traffic and inefficient when making many small stops. A newer idea has emerged to solve this: pairing a heavy-duty truck with a fleet of small, agile drones. In this system, the truck acts as a mobile base camp, driving along a main route while launching drones to zip directly to specific customers and then returning to the truck for a quick recharge or handoff. While the concept promises to slash costs and speed up deliveries, the mathematics of coordinating them is incredibly difficult. The truck and drones must move in perfect sync, respecting strict time windows for deliveries, battery limits, and the physical constraints of the vehicles, all while trying to minimize cost, time, and customer frustration simultaneously.
Researchers at Hebei University of Technology have tackled this complex coordination puzzle by developing a new computer method to find the best possible delivery plans. They created a mathematical model that treats the problem as a three-way balancing act: keeping costs low, finishing all deliveries as quickly as possible, and ensuring customers receive their packages at a time that suits them. Because there is no single "perfect" answer that wins on all three counts at once, their goal was to generate a wide variety of high-quality options, showing exactly how much faster a delivery could be if the company were willing to spend a little more money, or how much cheaper it could get if the customer accepted a slightly later arrival.
To navigate the massive number of possible combinations, the team designed an intelligent search algorithm called RC-NSGA-II. Imagine the search for the best route as exploring a vast, dark landscape where the goal is to find the highest peaks. Standard search methods often wander aimlessly or get stuck in small hills, missing the true mountains. This new algorithm works by dividing the landscape into distinct regions and keeping a memory of which areas have recently produced the best results. It then directs its computational energy toward those promising regions, much like a hiker who, after finding a few good views, decides to focus their search on that specific valley while still keeping an eye on the rest of the map to ensure nothing is missed. This "adaptive resource allocation" allows the system to learn as it goes, refining its search to find better solutions faster than previous methods.
The researchers tested their method on four different scenarios, ranging from small deliveries with twenty customers to large-scale operations with eighty customers. In these tests, their new algorithm consistently outperformed six other established methods used in the field. It found solutions that were not only better overall but also reached those high-quality results faster. For instance, on the largest test cases, it produced superior delivery plans in less than half the time required by a competing advanced method. The study also included a "what-if" analysis to help human managers make final decisions. By adjusting the priorities, the system could recommend a specific route for a company focused on saving money, another for one focused on speed, and a third for one focused on maximum customer satisfaction.
The results showed clear trade-offs in the real-world operations. When the priority was strictly cost, the system recommended a plan using only four drone flights, relying mostly on the truck, which kept expenses low but took longer. When speed was the goal, the system utilized all available drones for ten separate flights, cutting the total delivery time significantly but increasing the cost. When the focus was on customer satisfaction, the system again used ten flights but arranged them to ensure packages arrived within the most preferred time windows, minimizing the penalty for being too early or too late. The study confirms that while the math behind these decisions is complex, the outcome is a flexible toolkit that can adapt to different business needs.
The researchers acknowledge that their current model is a simulation based on specific assumptions, such as using straight-line distances for drones and simplified road distances for trucks, rather than the messy reality of city traffic and winding streets. They also note that their tests involved a single truck and a fixed number of drones, leaving the challenge of managing multiple trucks for future work. However, the core finding remains robust: by intelligently directing the search effort toward the most promising areas of the problem, it is possible to generate a rich set of optimized delivery plans that balance cost, time, and service quality far better than before. This approach offers a practical path forward for logistics companies looking to integrate drones into their fleets without getting lost in the complexity of the coordination required.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.