Research On Improved Algorithm For Driver Preview And Fusion Of Input Information

The applications of intelligent driver assistance for high-speed vehicle, such as collision warning system, ACC, lane departure system and so on, the obtain of the distance information between the host vehicle and the traffic in object is one of the key issues. The more accurate the preview algorithm, the more precise the safety assessment of intelligent vehicle will be. After research, we find that there was a defection of the driver model preview algorithm used in the safety evaluation of the traffic in object previously, so the algorithm needs to be improved: We will take the movement of traffic in object in the preview time into consideration. For the improvement of preview algorithm, we need to solve two problems: the estimations for the stage of traffic in object and the stage of the host vehicle.The ways of accessing information about the traffic in object are radar, sonar, vision and so on. We use the radar information more than other information in engineering. The information from radar is the relative position, velocity of the traffic in object under the carrier coordinate system which the radar is fixed on. The information used for traffic in object preview is the stage of it under the navigation coordinating system.The information sources used for the estimation of the stage of the host vehicle are GPS, INS, vehicle CAN, OD, gyro and so on. Now there is not a kind of sensor that can be fully functional yet. The navigation information transmitted to the host vehicle should be seamless, as accurate as possible. So we need to study on the fusion method of multi-source information.Therefore, the main content of the paper can be summarized as follows:First of all, we do research on the improvement of the driver preview algorithm. On the base of the hypothesis of steady preview and dynamic calibration, we use the planar rigid body kinematics to predict the curve trajectory of the center of mass, and finally achieve the goal of the preview of the host vehicle and traffic in object.Secondly, we estimate the stage of the traffic in object. Through the coordinated transformation of radar data under different times, we obtain the relative movement of the traffic in object under the same coordinating system, and then use the Kalman filter to get the position, velocity and acceleration estimation of the traffic in object. Through the point synthesis theorem of speed and acceleration, we determine the stage of the traffic in object under the navigation coordinating system.Thirdly, we do the fusion of multi-source information of the host vehicle. We describe the navigation coordinate system that commonly used, the conversion equation between them, and analyze the GPS, INS system error sources. We design the navigation solution for the host vehicle: According to the number of satellites can be observed and Position Geometric Dilution of Precision (PDOP), we choose the different way for fusion. We focus on the position, velocity error Kalman filter under INS/GPS navigation.Finally, we do simulation and experiment to verify the fusion algorithm for the stage of the host vehicle and the estimation algorithm for the stage of the traffic in object. We simulate the prediction of the curve trajectory of the center of mass to verify the feasibility of the prediction method. We compare the data used for the estimates of the traffic in object and the recorded data of the movement of the traffic in object to verify that the estimate algorithm of the traffic in object is feasible. Finally, through the angle step input simulation, we verify the correctness of the estimation algorithm for the host vehicle. We collect the multi-source information of the host vehicle and compare the fusion of different data to verify the fusion program is reasonable.In this paper, we try to explore the multi-source information fusion program of the host vehicle and the program of stage estimation and preview of the traffic in object. On the base of these, we improve the driver preview algorithm, and the algorithm will do more work for the applications of intelligent driver assistance of high-speed vehicle.