Cooperatively Localizing Multi-Target by Dynamic UAV Formation Equipped with Bearing-only Sensors

In recent years, the applications of UAV swarm equipped with electro-optical pods have become increasingly widespread. Electro-optical pods are passive sensors that can only measure bearing, but not distance. In practical scenarios of UAV swarm continuous tracking of multiple targets, the time-sensitive moving targets may move out of or suddenly into the field of view of UAV. Collaboratively localizing multiple time-sensitive targets based on bearing-only measurements is a significant challenge for UAV swarm. Traditional direction-finding cross location methods have significant errors under noisy conditions. Existing target location algorithms based on filtering require fixed UAV formation to track specific target, making them unsuitable for dynamic formations. In this paper, we model the target motion with a constant velocity and linearize the process of UAV swarm observing the positions of targets. Then we use the Extended Kalman Filter to estimate the position of targets based on the above target motion model and linearized observation model. The observation model allows for changes of the UAV formation tracking a specific target. Emulation experiments in the VICON environment and simulation experiments in the AirSim environment validate the applicability and accuracy of this method in scenarios of dynamic UAV formation cooperatively localizing multiple targets.