Systematic surveillance for UAVs: A feedforward iterative learning control approach

UAVs (unmanned air vehicles) have contributed greatly to situational awareness through surveillance missions. However, complete autonomy of a UAV has yet to be realized due to the lack of reliable onboard sensing capabilities. Current research at the Air Force pairs UAVs with Unattended Ground Sensors (UGS) to create a system that autonomously patrols, detects, and isolates intruders on a road network. During the patrol phase a UAV must visit each UGS, transitioning to the isolation phase if an intruder is detected. Optimizing the UAV flight plan during the patrol phase will lead to a faster response time and higher probability of capturing the intruder. The goal of this work is to investigate a learning-based approach that will enable more efficient and effective surveillance operations. In the proposed framework, techniques from adaptive feedforward iterative learning control and a region of attraction-based tracking approach have been used to optimize the UAV flight plan between surveillance flights. The proposed approach has resulted in the development of a novel learning control framework that leverages a region-based tracking requirement to minimize overall distance traveled, while guaranteeing convergence within the required tracking zone. Simulation results for a 1D example system demonstrate the validity of the control framework through an 8% reduction of overall distance traveled as compared to traditional surveillance strategies requiring tracking convergence to a single point. The proposed framework has the potential to significantly decrease the resources required for surveillance-based applications.