Distributed collaborative localization method for UAV swarms using ground-based lidar point cloud

Abstract

A distributed collaborative positioning method for drone swarms, assisted by ground-based lidar point clouds, includes: drones performing initial pose estimation using a monocular camera and an inertial measurement unit (IMU); unmanned vehicles using the Fast-LIO algorithm to construct a dense point cloud map using lidar and IMU data; a factor graph optimization problem is constructed by combining sparse and dense point cloud data; each drone receives environmental information to achieve six-degree-of-freedom pose estimation; the drones independently run pose estimation modules and use the BA algorithm to accurately solve pose; and information synchronization is achieved between the drones and unmanned vehicles. This method achieves distributed, no-prior-required, real-time pose estimation for drone swarms and unmanned vehicles in unstructured environments. The distributed architecture enables multiple drones to independently solve their real-time poses based on a shared lidar map. The visual odometry factor graph, constructed by integrating lidar point cloud information, can reduce the drift problem of visual-inertial odometry over long trajectories.

Images