Terrain-based Vehicle Localization Using Low Cost MEMS-IMU Sensors

This paper presents a method for localizing a land vehicle using low cost MEMS IMU sensors and a terrain map in global positioning system (GPS) denied environment. Previous attempts to localize a vehicle based on terrain information used a tactical grade IMU which is not cost-effective on a commercial scale. On the other hand, MEMS sensors are very cheap but cannot provide accurate attitude estimation for a moving vehicle as the sensor measurements are influenced by the kinematic motion of the vehicle. Acceleration and braking corrupt the longitudinal sensor measurements and turning corrupts the lateral sensor measurements. In this paper, we formulate a Kalman filter which is able to compensate the external acceleration acting in the lateral direction using a gyroscope and odometery providing an accurate estimate of the vehicle's roll angle. A particle filter then uses this roll angle estimate along with a pre-determined terrain map of the driving region to localize the vehicle. Experimental results performed on an instrumented test vehicle show that the corrected roll measurements from MEMS IMU data using the proposed Kalman filter can localize a land vehicle very accurately.