Improving the performance of out-of-order sigma-point Kalman filters for joint localization by multiple UAVs

In search and surveillance operations, a team of small Unmanned Aerial Vehicles (UAVs) can provide a robust solution that outperforms what can be achieved by a single aircraft with comparatively superior mobility and sensors. Three key challenges can be identified in this field: (1) optimizing the UAV trajectories to place them at desired locations at desired times to capture target locations, (2) cooperative sensor scheduling, and (3) intelligent fusing of multiple sensor measurements to accurately estimate the position and velocity of a target. The focus of this paper is the sensor-fusion task, in particular with the goal of providing accurate localization of a ground target from coarse azimuth angle-of-arrival information captured by multiple cameras onboard flying platforms. One might consider addressing this problem using some form of Kalman filter. However, the estimation system is nonlinear and, in practical implementation, sensor readings can arrive out-ofsequence to the sensor fusion process. For example, there is non-deterministic latency in the inter- and intra-UAV communication channels. We have previously addressed both issues by developing an out-of-order sigma-point Kalman filter (O3SPKF) [1], and in this paper we seek to improve on its localization performance by coupling O3SPKF with the triangulation of the azimuth angles simultaneously registered by the cameras in multiple UAVs.