Locker-based Truck-Drone Routing with Integrated Considerations of Pickups, Deliveries, and No-Fly Zones

The Logistics Puzzle: When Drones, Trucks, and Lockers Collide

A new preprint on arXiv (2606.30680v1) tackles a genuinely practical problem in last-mile logistics: how to route a hybrid fleet of trucks and drones when smart lockers serve as intermediate nodes, and no-fly zones constrain aerial paths. The research formalizes a "locker-based truck-drone routing problem" that simultaneously optimizes pickups, deliveries, and drone flight restrictions—a combinatorial challenge that goes well beyond the standard traveling salesman with drones.

What the Research Actually Does

The paper extends existing truck-drone routing models by integrating three real-world constraints that are often treated separately. First, lockers act as temporary parcel storage, meaning a truck can drop off packages at a locker, then a drone can later retrieve and deliver them—or vice versa. This decouples the timing of truck and drone operations, introducing complex synchronization requirements. Second, the model handles both pickups and deliveries simultaneously, which is more realistic than one-way delivery-only scenarios. Third, no-fly zones (e.g., airports, military installations, dense urban cores) are encoded as hard constraints on drone paths, forcing the algorithm to find feasible aerial routes that avoid restricted airspace.

The authors propose a mixed-integer programming formulation and likely a heuristic or metaheuristic solution approach (the abstract is not fully detailed here, but the problem's NP-hard nature suggests exact methods will scale poorly). The key contribution is the unified framework that forces the optimizer to trade off truck travel costs, drone flight times, locker capacity, and no-fly zone avoidance simultaneously.

Why This Matters for Last-Mile Logistics

The practical significance is immediate. Companies like Amazon, UPS, and Wing have all experimented with truck-drone hybrid systems, but operationalizing them at scale requires solving exactly this kind of integrated routing problem. Smart lockers are already proliferating in apartment buildings and retail locations; combining them with drone delivery could dramatically reduce last-mile costs and delivery times. However, no-fly zones are a persistent regulatory reality—drones cannot simply fly point-to-point over sensitive areas. This research provides a mathematical foundation for planning around those restrictions rather than ignoring them.

For logistics operators, the model offers a way to evaluate trade-offs: Is it cheaper to send a truck around a no-fly zone or to use a drone that flies over it? Should a locker be used as a relay point or a final delivery destination? The answers depend on the specific cost parameters, but having a formal optimization framework means these decisions can be data-driven rather than heuristic.

Implications for AI Practitioners

For those building AI systems for logistics, this paper highlights several design principles:

• Constraint integration matters. Separating routing, locker assignment, and no-fly zone avoidance into independent modules will produce suboptimal solutions. The best results come from a unified optimization that captures interactions—e.g., a drone's ability to avoid a no-fly zone depends on where the truck drops it off.

• Scalability is the real challenge. Exact methods will likely fail for city-scale instances with hundreds of delivery points. Practitioners should focus on developing efficient heuristics, perhaps using reinforcement learning or large neighborhood search, that can approximate the optimal solution in seconds rather than hours.

• Real-world data is essential. The paper's value depends on realistic cost parameters (truck fuel, drone battery, locker rental) and no-fly zone geometries. AI practitioners should push for open benchmarks that include actual city maps, traffic patterns, and airspace restrictions.

Key Takeaways

• The paper unifies three previously separate logistics constraints—locker-based operations, simultaneous pickups/deliveries, and no-fly zones—into a single optimization model.
• This integration is crucial for practical truck-drone delivery systems, where ignoring any one constraint leads to infeasible or inefficient routes.
• AI practitioners should prioritize scalable heuristic solvers over exact methods, as the problem is NP-hard and real-world instances will be large.
• The research underscores the need for realistic benchmarks with actual city data, no-fly zone maps, and cost parameters to drive further progress in autonomous last-mile logistics.