Research On Relevant Target Selection Algorithm Of Adaptive Cruise Control System

Adaptive cruise control (ACC) system is an extension of conventional cruise control system, which has the ability to cruise at reference speed when there is no relevant target forward. This system can detect objects before the host vehicle by range sensor like radar, select relevant target, and automatically control vehicle longitudinal velocity to keep an appropriate safe headway distance between the host vehicle and the relevant target through actuating throttle or brake. In this way, the system can lessen driver's workload, and keep driving more comfort and safety.In complicated multi-vehicle traffic scenarioes, the object that the host vehicle following is called relevant target.The relevant target selection algorithm plays crucial role in the ACC system. Only when the relevant target is selected or facts show that there is no relevant target, the controller of ACC could start to work. Any mistake may cause serious consequence. For example, being blind to a slower vehicle in the same lane may occur serious collision. In the other case, mistaking the object in the adjacent lane as relevant target may affect passing speed.In this thesis, a tracking-based relevant target selection algorithm of vehicle adaptive cruise control was researched. First, the history of ACC and some previous research of our workteam on this area were reviewed, and then the critical role of relevant target selection and several related algorithms were analyzed. Next, the tracking-based relevant target selection algorithm was theoretically explained. Through radar information and host vehicle states, a kalman filter was used to estimate each object's states referring to the host. From these states, road curvature and host path angle can be extracted. Then through estimation of host lane lateral position relevant target was selected. Then, based on co-simulation of Carsim&Simulink, the algorithm was validated. More accurate estimation of host future driving course and the situation where stationary objects exist was discussed. Finally, a data acquisition experimental platform was developed. The data characteristics of sensors like radar, gyroscope and non-contact optical sensor were analyzed. Some basic experiments were conducted outside with this platform. The main contents and conclusions are as follows:(1) Traditional constant curvature relevant target selection algorithm has many drawbacks in transition road segment like curve-entry, curve-exit and lanechange. To some extent, the observed trajectories of preceding objects could help to determine the shape of the road in front of the host. Along with the path angle of the host w.r.t. the road, the relevant target would be accurately selected.(2) Through continuously tracking each preceding object, all the objects' motion states can be estimated, along with host vehicle states, the road before and future host driving course can be determined. by virtue of lateral offset referring to the host lane to judge whether each object is in the same lane with the host or not. Then the clostest object in the same lane is chosen as the relevant target. This thesis uses kalman filter to estimate object states, road curvature and host path angle etc. In later co-simulation of Carsim&Simulink, traffic scenarioes like straight, curve, curve-entry, curve-exit, object lanechange were simulated, and the results showed that the tracking-based relevant target selection algorithm was valid in all these scenarioes.(3) A data acquisition experimental platform was developed on vehicle Jetta. The hardware platform includes various sensors, data collection cards and industrial computer. The software platform is based on NI LabVIEW, on which the acquisition system was developed. Delphi ESR radar was mounted on the center of the front bumper, which would return objects' range, range rate and azimuth. Gyroscope sensor was mounted on the CG of host vehicle, which could get host yawrate. Non-contact optical sensor was mounted on the rear of host vehicle, which could get the host longitudinal and lateral velocity. Finally, two experiments were conducted in straight and circle conditions, which are helpful for further experimental researches on relevant target selection of vehicle adaptive cruise control.