Multi-Objective Fox Optimization with Adaptive Kalman Filtering for UAV Autonomous Navigation

Abstract

A Multi-objective Fox Optimization Algorithm coupled with an Adaptive Kalman filter was proposed in this paper to build an improved autonomous navigation framework to Unmanned Aerial Vehicles (UAVs). The method addresses key challenges with respect to the real-time path planning, obstacle avoidance, and environmental properties. Adaptive Kalman Filter improves the fusion of multi-sensor data attributes due to the dynamically tuning covariance by limiting the noise factor which results in the forecasting of more precise states. To optimise the navigation routes, the Multi-objective Fox Optimization Algorithm balances path length, energy efficiency, and the obstacle safe margins. The simulator experiments were done under AirSim indoor (100x100m) and outdoor (250x250m) settings with three obstacle densities, simulated multi-sensor data (LiDAR 20 Hz, RGB 30 FPS, GPS/IMU 50 Hz). The trained method was evaluated over 30 independent runs of each scenario and trained on 3 000 trajectories (70/15/15 split). They demonstrate that the results have an average path-planning accuracy of 95.5 % (1.2) with a high runtime and energy efficiency improvement compared to state-of-the-art baselines. The system is developed in a well-structured pipeline process that consists of data collection, preprocessing, model training, decision model generation, optimisation, and deployment using hardware-in-the-loop simulation. Performance evaluation pits the suggested approach against the A* algorithm, the Particle Swarm Optimization (PSO) algorithm, and the Genetic Algorithm (GA) in accuracy, precision, recall, and F1-score. As the experimental results demonstrate, the proposed method reaches 95.53 percent accuracy, 94.15 percent precision, 93.62 percent recall, and 93.87 F1-score and is better than all baselines. All these improvements are significant according to statistical t-tests (p < 0.001), and the reduced variance shows higher robustness and reliability in a wider range of application. The proposed approach to monitoring UAVs in the complicated environment shows great potential in real-world UAV missions, including environmental monitoring, disaster response, observation, and logistics. The use of adaptive sensing combined with multi-objective optimization brings a tradeoff between accuracy and efficiency that will guarantee resilience in autonomous UAV navigation.