Autonomous Tracking and Terminal Guidance of Moving Targets for Fixed-Wing UAVs

This paper presents an integrated autonomous control framework designed for fixed-wing unmanned aerial vehicles (UAVs) equipped with pan-tilt (PT) cameras. The system is engineered to manage the entire mission lifecycle, from initial target acquisition and tracking to final terminal engagement. By combining vision-based estimation with advanced control laws, the authors address the inherent difficulties of tracking moving targets with fixed-wing platforms, such as limited maneuverability, gimbal constraints, and the risk of visual self-occlusion during banking maneuvers.

The proposed system operates in three distinct phases:

1. Target Acquisition and Estimation: The system uses a YOLO-based detection algorithm to identify targets. An Unscented Kalman Filter (UKF) fuses these visual detections with inertial measurements to estimate the target's state, even when the target's motion dynamics are unknown.
2. NMPC-based Tracking: A Nonlinear Model Predictive Control (NMPC) strategy manages the UAV's flight path. This controller incorporates Control Barrier Functions (CBFs) to explicitly prevent the UAV from obstructing its own line-of-sight (self-occlusion) during aggressive turns, while respecting the physical limits of the airframe and the PT camera.
3. Terminal Guidance: Upon reaching engagement conditions, the system transitions to a quaternion-based Biased Proportional Navigation Guidance (BPNG) law. This ensures the UAV achieves a precise impact angle on the target without requiring complex manual gain tuning.

Fixed-wing UAVs are highly efficient for long-endurance missions, but their flight dynamics make precise tracking and terminal engagement significantly more difficult than for rotor-based platforms. This research provides a robust, unified solution that bridges the gap between perception and control. By explicitly modeling the interaction between the UAV's bank angle, gimbal constraints, and the target's relative position, the framework enables reliable, autonomous operations in complex environments where visual contact must be maintained throughout the entire engagement sequence.