The application of drones in last-mile distribution has been a contentious research topic in recent years. Existing urban distribution modes mostly depend on trucks. This paper proposes a novel package pickup and delivery mode and system wherein multiple drones collaborate with automatic devices. The proposed mode uses free areas on top of residential buildings to set automatic devices as delivery and pickup points of packages, and employs drones to transport packages between buildings and depots. The integrated scheduling problem of package drop-pickup considering m -drones, m -depots, and m -customers is crucial for the system. Therefore, we propose a simulated-annealing-based two-phase optimization (SATO) approach to solve this problem. In the first phase, tasks are allocated to depots for serving, such that the initial problem is decomposed into multiple single-depot scheduling problems with m -drone. In the second phase, considering the drone capability and task demand constraints, we generated a route-planning scheme for drones in each depot. Concurrently, an improved variable neighborhood descent (IVND) algorithm was designed in the first phase to reallocate tasks, and a local search (LS) algorithm was proposed to search for high-quality solutions in the second phase. Finally, extensive experiments and comparative studies were conducted to verify the effectiveness of the proposed approach.

Unmanned aerial vehicles (UAVs) have emerged as promising platforms for fast, energy-efficient, and cost-effective package delivery. Path planning in 3-D urban environments is critical to drone delivery. The paper proposes a novel tangent-based (3D-TG) method for UAV path planning in 3-D urban environments. When a drone encounters an obstacle, a tangent graph is constructed to generate three sub-paths from both sides and above to bypass an obstacle, one of which is selected according to sophistically designed heuristic rules. The selected sub-path would be constantly adjusted its direction via tangent graph to avoid obstacles until the path can extend to the goal without obstacle collision. To avoid moving obstacles, velocity obstacle is incorporated in the 3D-TG. The experimental results on synthetic and realistic scenarios illustrate that 3D-TG performs well under static, unknown and dynamic environments. More significantly, 3D-TG can also generate a collision-free path for a drone to navigate through simple mazes efficiently, within a reasonable time.

Unmanned aerial vehicles are becoming promising platforms for disaster relief, such as providing emergency communication services in wireless sensor networks, delivering some living supplies, and mapping for disaster recovery. Dynamic task scheduling plays a very critical role in coping with emergent tasks. To solve the multi-UAV dynamic task scheduling, this paper constructs a multi-constraint mathematical model for multi-UAV dynamic task scheduling, involving task demands and platform capabilities. Three objectives are considered, which are to maximize the total profit of scheduled tasks, to minimize the time consumption, and to balance the number of scheduled tasks for multiple UAVs. The multi-objective problem is converted into single-objective optimization via the weighted sum method. Then, a novel dynamic task scheduling method based on a hybrid contract net protocol is proposed, including a buy-sell contract, swap contract, and replacement contract. Finally, extensive simulations are conducted under three scenarios with emergency tasks, pop-up obstacles, and platform failure to verify the superiority of the proposed method.

Existing neural heuristics for multiobjective vehicle routing problems (MOVRPs) are primarily conditioned on instance context, which failed to appropriately exploit preference and problem size, thus holding back the performance. To thoroughly unleash the potential, we propose a novel conditional neural heuristic (CNH) that fully leverages the instance context, preference, and size with an encoder–decoder structured policy network. Particularly, in our CNH, we design a dual-attention-based encoder to relate preferences and instance contexts, so as to better capture their joint effect on approximating the exact Pareto front (PF). We also design a size-aware decoder based on the sinusoidal encoding to explicitly incorporate the problem size into the embedding, so that a single trained model could better solve instances of various scales. Besides, we customize the REINFORCE algorithm to train the neural heuristic by leveraging stochastic preferences (SPs), which further enhances the training performance. Extensive experimental results on random and benchmark instances reveal that our CNH could achieve favorable approximation to the whole PF with higher hypervolume (HV) and lower optimality gap (Gap) than those of the existing neural and conventional heuristics. More importantly, a single trained model of our CNH can outperform other neural heuristics that are exclusively trained on each size. In addition, the effectiveness of the key designs is also verified through ablation studies.

Efficient scheduling is critical for the effective use of heterogeneous UAVs equipped with various sensors. The collaborative “electronic signal guided imaging” reconnaissance mode is first discussed in this paper, where electronic signal reconnaissance equipment is initially used to locate electromagnetic targets over long distances, offering approximate target locations that can guide the next imaging reconnaissance. We construct a multi-objective optimization model for heterogeneous multi-UAV coordinated scheduling that minimizes the total travel cost and maximizes the overall weight of scheduled tasks simultaneously. A local search-based multi-objective algorithm, called multi-objective variable neighborhood descent (MOVND), is proposed to solve the problem. Specifically, a customized variable neighborhood descent is designed to optimize each objective in parallel, which contains several sophisticatedly designed problem-specific heuristics selected by adaptive weight adjustment. Additionally, Metropolis acceptance criteria, which employ adaptive temperature control, are incorporated to prevent premature convergence and improve optimization capability. Furthermore, knee solutions and boundary solutions are optimized specifically to further enhance the optimization. Various local search strategies for two objectives are implemented in an iterative manner until the predefined stopping criteria are satisfied. The proposed algorithm is tested on 18 benchmark instances and one real-world instance. The experimental results on benchmark instances show the superiority of our proposed algorithm with regard to diversity and convergence compared with seven effective algorithms, especially for large-scale instances. The effectiveness and practicability of our proposed algorithm are verified by an additional result on a real-world instance.