Research on Low-Altitude Flight Obstacle Avoidance System for Police UAV Based on Multi-Sensor Fusion

Police uncrewed aerial vehicles (UAVs) operating in complex low-altitude urban environments face a fundamental trade-off between dynamic obstacle avoidance and covert penetration under strong electromagnetic interference and dense structural occlusion. Conventional single-sensor or fixed-weight fusion schemes exhibit severe performance degradation under such conditions due to spatiotemporal misalignment and sensor reliability fluctuations. To address these challenges, this paper proposes a confidence-weighted multi-sensor fusion framework that dynamically adapts sensor contributions according to real-time interference intensity and motion disturbance. A three-stage adaptive mechanism is developed, integrating environmental interference assessment, sensor reliability feedback, and incremental weight optimization to mitigate cross-modal data conflicts in highly dynamic scenarios. To further enhance perception consistency, a Lie-group-based spatiotemporal joint calibration method is introduced, which compensates for clock drift and spatial misalignment among heterogeneous sensors, significantly improving trajectory planning accuracy under urban canyon conditions. On this basis, a concealment–obstacle avoidance cooperative decision-making strategy is formulated as a dual-objective dynamic game and solved using an improved Nash-Q algorithm. By incorporating dynamic learning-rate decay and Pareto-optimal constraints, the proposed algorithm achieves stable convergence to an $\varepsilon $ -equilibrium while avoiding oscillation commonly observed in multi-objective reinforcement learning.Experimental results under extreme electromagnetic interference and adverse environmental conditions demonstrate that the proposed framework effectively reduces cross-modal fusion errors and collision probability while maintaining low observability. Compared with conventional fusion and control strategies, the system achieves faster decision convergence and improved trajectory accuracy, validating its robustness and applicability for police UAV missions in complex urban environments. The proposed approach provides a unified perception–decision framework for reliable low-altitude UAV operation under strong interference and concealment constraints.