Cascaded Kalman Filters for Accurate Estimation of Multiple Biases, Dead-Reckoning Navigation, and Full State Feedback Control of Ground Vehicles

Abstract

This paper develops a cascaded estimation algorithm for estimating all of the biases and states for full state feedback and dead reckoning of a farm tractor through short global positioning system (GPS) outages. First, a conventional (one stage) estimation scheme is presented. The single state estimation scheme is shown to have degraded performance in bias state estimation and dead-reckoning due to vehicle model errors. However, the states for position and velocity are not highly coupled to the tractor dynamic states, allowing for separation of the estimators. Therefore, the state estimation algorithms are divided into two cascaded estimators in order to prevent the errors in the vehicle model from corrupting the navigation states. A dead reckoning (or navigation) estimator estimates all of the inertial sensor biases while GPS is available. When GPS is not available, the dead reckoning estimator integrates rate measurements to provide position and heading estimates in order to maintain continuous control of the vehicle through these GPS outages. A second estimator is then used to estimate the additional states needed for full state feedback control algorithms. Bias estimates from the dead reckoning estimator are used to correct the sensor measurement used in the second estimator. An extended kalman filter (EKF) is utilized for each of the estimators. Results are given, showing that the cascaded estimation technique provides better estimation of the vehicle states over a conventional estimation scheme, especially during a GPS outage. Results are also given which verify the ability of the estimation algorithm to estimate all of the system biases and provide continuous control of the tractor through a short GPS outage