Robust estimation with faulty measurements using recursive-RANSAC

Many autonomous platforms, such as micro air-vehicles, are increasingly relying on cheap, lightweight sensors to improve the low-level state estimation for navigation and control. Unfortunately, these and all sensors have a finite probability of returning spurious measurements that do not follow the classical zero-mean Gaussian models of measurement noise. A classical heuristic used to mitigate the effects of sensor faults is the gated-Kalman filter. We show that the gated- Kalman filter estimate diverges from the true states when the probability of detection is low or when the measurement noise standard deviation increases above the expected value. The main contribution of this paper is to utilize the recently developed recursive-RANSAC algorithm in a feedback loop to robustly estimate the true states when the probability of a sensor fault is high, when the measurement noise characteristics abruptly change, and during brief occlusions of the true signal, while maintaining real-time performance.