Image-Processing-Based State Estimation for Lateral Control of Electric Vehicles Using Multi-Rate Kalman Filter

Driver assistance systems such as automatic steering for lane keeping are of particular importance for vehicle's lateral safety, and onboard look-ahead cameras are widely employed for realization of such applications. In fact, many papers and patents on vision-based lane keeping and automatic steering have been published in the past few decades. On the other hand, electric vehicle (EV) as a green solution for future transportation is gaining attentions nowadays, and the differential torque between the left and right wheels is considered as an effective method for their lateral safety control. Nevertheless, the sampling rate of normal cameras is 30 Hz, which is much slower compared with that of motors and other kinds of onboard sensors, and the time delay caused by image processing also needs to be considered. Previous researches and patents simply adapt the whole system's sampling frequency to that of the slowest device; however, the held and delayed feedbacks deteriorate system performance and may cause instability. In this paper, the two problems are solved using a multi-rate Kalman filter (KF) with measurement delays included. Based on our experimental EV, the performance of the proposed multi-rate Kalman filter is verified with both simulations and experiments.