Anomaly Resilient Relative Pose Estimation for Multiple Nonholonomic Mobile Robot Systems

This article addresses the resilient relative pose estimation problem for multiple mobile robot systems against abnormal sensor measurements. Motivated by the fact that in real implementations, sensors used for neighboring robot detection, such as stereo camera, laser range finder, etc., may suffer from unpredictable anomalies, a resilient relative pose estimation approach is proposed such that each robot can obtain satisfactory relative pose estimates of its neighbors for further coordination algorithm design. In the proposed approach, the optimal Kalman estimate is decomposed as a weighted sum of local state estimates, based on which a convex optimization problem is formulated to generate the resilient estimation. Unlike most of the existing approaches investigating similar problems, which assume that the statistics or the bound of anomalies is known in advance, our proposed approach is not limited by this assumption. The effectiveness of the proposed method has been validated by numerical simulations and real robot experiments. It has been demonstrated that the proposed approach is comparable to the Kalman filter in the absence of sensor anomalies, while boundedness of the relative pose estimation error can be guaranteed under abnormal observations.