Multi-rate Kalman Filter Design for Electric Vehicles Control based on Onboard Vision System with Uneven Time Delay

Abstract

Vehicle safety systems have become increasingly popular in modern vehicles, and real time information of vehicle states such as yaw rate, lateral acceleration and lateral position is indispensable for such systems. Yaw rate and acceleration signals can be obtained from onboard gyro/inertial sensors, and vehicle lateral position can be measured by an onboard vision system. Normally, the sampling rate of a camera is much slower compared with that of the other onboard sensors. Moreover, image processing takes time and the time varies depending on captured images and hardware loads (delay time is measurable through time stamp). In case of integrated vehicle motion and position control, however, a unified feedback frequency is desired. Considering the slow control periods of traditional actuators such as hydraulic brakes, many previous studies down-sample the fast rate sensors to adapt the vision device. On the other hand, for electric vehicles, the control period of actuators (motors) is much shorter than the sampling time of a normal camera. To improve the control performance, this research employs a combined vehicle and vision model for lateral position estimation and proposes a multi-rate Kalman filter with reconstructed measurements based on an inter-sample residual estimation technique.