A Cooperative UAV Swarm Localization Algorithm Based on Probabilistic Data Association for Visual Measurement

A cooperative unmanned aerial vehicle (UAV) swarm has been proven to be a fast and efficient system for performing tasks. The relative measurements based on visual detection will cause data association ambiguity in a swarm with homogeneous visual appearance. To solve this problem, this article introduces a UAV swarm localization system based on probabilistic data association (PDA) for vision-based relative measurements. First, a cooperative localization system is designed that each UAV in the swarm shares its information with other UAVs to help the UAV without global navigation satellite system (GNSS) achieve robust localization. The effect of the position uncertainty of assisted UAVs on the localization performance is considered in the process of visual measurement. Then, this work uses the modified directional joint PDA (MDJPDA) algorithm to solve the problem of data association for visual measurement in the homogeneous UAV swarm. By adding the directional weighting factor, the localization errors caused by the visual observation angle are reduced. Finally, the simulation and experimental results demonstrate that the proposed algorithm is superior to the extended Kalman filter (EKF) algorithm, the modified PDA (MPDA) algorithm, and the modified joint PDA (MJPDA) algorithm in terms of localization accuracy and robustness.