Applying Cooperative Localization to Swarm UAVS Using an Extended Kalman Filter.

摘要

Cooperative Localization (CL) is a process by which autonomous vehicles operating as a team estimate the position of one another to compensate for errors in the positioning sensors used by a single agent. By combining independent measurements originating from members of the team, a single estimate of increased accuracy will result. This approach has the potential to enhance the positional accuracy of an agent over use of a standard GPS, which would be essential for behaviors within a swarm requiring precision movements such as maintaining close formation. CL can also provide accurate positional information to the entire group when operating in an intermittent or denied GPS environment. In this thesis, a distributed CL algorithm is implemented on a swarm of Unmanned Aerial Vehicles (UAVs) using an Extended Kalman Filter. Using a technique created for ground robots, the equations are modified to adapt the algorithm to aerial vehicles, and then operation of the algorithm is demonstrated in a centralized system using AR Drones and the Robot Operating System. During tests, the positional accuracy of the UAV using CL improved over use of dead reckoning. However, the performance is not as expected based on the results noted from the referenced two-dimensional application of the algorithm. It is presumed that the sensors on-board the AR Drone are responsible. Since the platform is simply a low-cost solution to show proof-of-concept, it is concluded that the implementation of CL presented in this thesis is a suitable approach for enhancing positional accuracy of UAVs within a swarm.