Interacting Multiple Model Kalman Filter based Vehicle Lateral Motion Estimation under Various Road Surface Conditions

Abstract

In this paper, we present an estimation method of vehicle lateral motion using interacting multiple model (IMM) Kalman filter (KF) to consider various road surface conditions for the vehicle driving on asphalt, wet, or snow road. In a vehicle lateral dynamic model, the exact values of cornering stiffness are unknown so that typical nominal values are used for obtaining the control law for the active safety systems. To cope with this problem, in this paper we propose the IMM which consists of two vehicle dynamic lateral motion models: One model has a nominal dry road parameter, and the other one does parameters regarding the snow road. From IMM KF, we can obtain the stochastically best-blended state of the vehicle over various road surface conditions. From the simulation results on real measured camera sensor data, we observed that the vehicle lateral motion estimation using IMM KF outperforms the estimation using KF of each model. Furthermore, we validated the effectiveness of the proposed method using the virtual lane with the output of IMM KF estimates under various road surface conditions.