Intent-informed UAV tracking for evasive targets in urban environments

The utilization of unmanned aerial vehicles (UAVs) for the search and tracking of targets within urban environments is a well-researched area with broad applications. However, there has been limited research addressing the challenge of tracking targets exhibiting evasive behavior. Such targets can escape the UAV's tracking through adversarial movements. In response to this issue, the present study introduces a comprehensive framework for target tracking that integrates inference of the target's intentions with trajectory prediction. In this framework, the target's intention is characterized by potential goal regions, and the trajectory data of the observed target is utilized to infer these intentions within a Bayesian context. Conditioned on the inferred intention, the target's future motion is predicted through sampling. The resulting trajectory forecast underpins a planning method that seeks to maximize the cumulative future detection probability. Simulation results indicate that inference of the target's intentions significantly enhances the UAV's search efficacy and improves overall target tracking performance.