GNSS spoofing can progressively corrupt unmanned aerial vehicle (UAV) navigation when manipulated position solutions are fused before integrity assessment. Existing detection methods often use dead-reckoning references such as visual–inertial odometry (VIO) without fully accounting for accumulated drift or GNSS-dependent global alignment. This paper presents a closed-loop software-in-the-loop (SITL) evaluation platform and a detection-aware delayed multi-sensor fusion framework for UAV GNSS spoofing mitigation. The platform couples state estimation, flight control, vehicle dynamics, and sensor feedback to enable Monte Carlo evaluation under realistic navigation–control interactions. A cumulative sum (CUSUM) detector monitors consistency between GNSS and drift-compensated VIO, while recent GNSS factors are quarantined and incorporated into a factor graph only after integrity screening. Experiments quantify false-alarm rate, detection probability, and latency across detector settings and spoofing deviation rates. Onboard implementation and controlled injection experiments further assess real-time feasibility and hardware transferability. Under the evaluated sensor conditions and gradual-drift attack model, a 10–15 s window yielded detection probabilities exceeding 96% for attacks with a deviation rate of 0.5 m/s, with an approximately 3% false-alarm rate. These findings demonstrate the framework’s potential for spoofing mitigation and the platform’s utility for closed-loop evaluation.

Coordinating unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs) for post-disaster emergency material distribution requires joint decisions on task allocation, routing, delivery quantities, and rendezvous scheduling. Mobile UGV bases and split deliveries increase operational flexibility but introduce strong dependencies between aerial and ground operations. This paper investigates a centralized multi-objective multi-robot task allocation problem with static, known demands, in which UGVs transport supplies and support UAV launch, recovery, and battery replacement. A three-objective model is formulated to minimize weighted demand completion time, operating cost, and maximum normalized workload, subject to accessibility, payload, inventory, energy, and synchronization constraints. To solve the resulting constrained combinatorial optimization problem, we propose HV-ADEA, a hypervolume-guided adaptive evolutionary algorithm. A hybrid representation jointly encodes UGV stop sequences, UAV–interval task assignments, and split delivery quantities, while a synchronized decoder derives execution times and evaluates feasibility. A bounded iterative repair mechanism addresses coupled constraint violations. Domain-specific variation operators perform joint mobile base–sortie adjustment, aerial–ground quantity transfer, delivery splitting and merging, and synchronization bottleneck reduction to explore alternative allocation and routing plans. An experimental protocol is established using six test configurations of increasing scale and five baseline algorithms: NSGA-II, CCMO, BiCo, EMCMO, and C3M. The test results show that the algorithm we propose is competitive in most of the test cases.

Graph neural networks (GNNs) are highly effective on relational data, yet their performance degrades sharply when graph topology is poisoned before training. Existing defenses usually assume a fixed attack pattern and a fixed graph structure, which makes them brittle when the poisoned graph changes across attacks, perturbation budgets, or deployment conditions. We propose GADD, a game-inspired adversarial distillation framework for robust graph defense. GADD first constructs multiple positive and negative graph views through a homophily-aware graph sampling scheme, allowing the model to learn from both purified and high-risk subgraphs. It then trains a heterogeneous group of student GNNs online, where each student receives global class-distribution knowledge from its peers and local structural knowledge through an adversarial cyclic distillation objective. Finally, GADD replaces uniform ensembling with an entropy-regularized adaptive aggregation rule that assigns graph-adaptive weights according to confidence and inter-model agreement. On Cora, CiteSeer, and PubMed, GADD consistently improves robustness against both Meta and Nettack attacks while preserving clean accuracy. Under the strongest Meta and Nettack settings in the main benchmark, GADD improves the best competing baseline by up to 2.99 and 3.42 percentage points, respectively. Additional ablations show that graph sampling, adversarial distillation, and adaptive aggregation all contribute materially to the final robustness gains.