Author: Anonymous

Two prediction approaches are compared for a UAV’s altitude…

Two prediction approaches are compared for a UAV’s altitude channel. Model 1 predicts the next altitude reading as a weighted combination of recent altitude readings. Model 2 predicts it from the vehicle’s state and the thrust commands the controller issued. An adversary begins slowly biasing the barometer while the controller commands a hold. Which model is better positioned to detect this, and why?

An EKF tracking a UAV’s attitude is initialized with a poor…

An EKF tracking a UAV’s attitude is initialized with a poor estimate. Over the next several seconds the estimate drifts further from truth while the reported covariance shrinks. A linear filter with a comparably bad initialization would have recovered. What accounts for the difference?