Autonomous Aerial Interception in Cluttered Environments: An Integrated Perception–Guidance Framework With Adaptive State Estimation

Abstract

Autonomous kinetic interception of noncooperative low, slow, and small aerial targets in cluttered environments poses a fundamental challenge to traditional guidance and control systems. This task demands a rigorous balance between aggressive pursuit maneuvers and strict collision avoidance constraints. Existing methods typically decouple target tracking from trajectory planning, neglecting the coupling effects between target maneuvers and environmental obstacles. To address this, an integrated perception–guidance–control framework is proposed for aerial interception. The system introduces three primary contributions to the field of autonomous guidance: First, robust target state estimation: An innovation-based adaptive interacting multiple model estimator is developed to track maneuvering targets. By dynamically adjusting the process noise covariance matrix based on the real-time innovation sequence, the filter significantly reduces tracking latency during abrupt target acceleration. Second, guidance-integrated trajectory optimization: A tactical planning module that incorporates the classical proportional navigation guidance law into a gradient-based trajectory optimizer (MINCO) is developed. This approach generates dynamically feasible interception paths that conform to a time-dependent safety radius, explicitly accounting for perception uncertainty. Finally, reactive safety mechanism: A repulsion-field-based active evasion strategy is implemented to resolve local minima deadlocks, ensuring the interceptor’s survivability in dense obstacle fields. The effectiveness of the proposed framework is verified via comprehensive Monte Carlo simulations and real-world flight tests on an embedded computing platform. Results demonstrate that the system achieves high-precision physical interception and robust obstacle avoidance against dynamic targets in complex environments.