Research On Key Technology Of Control And Decision In Driver Assistant System

As for the traffic accident caused by drivers, there are two traditional solutions: autonomous driving technology and driver assistant system (DAS). However, complete auto-driving application depends on a whole intelligent transportation system (ITS), relative laws and human ethics, so it's hard to be practical in short term. Most existing DAS are warning-oriented, lacking necessary interference in emergency situations. To solve this problem, a"dual-mode"DAS is proposed which could take advantage of vehicle actively in emergency situations. It could switch between human-mode and auto-mode intelligently according to driving environment. In auto-mode, computer will control vehicle to assure driving safety. A prototype vehicle of this"dual-mode"is invented; active collision avoidance and lane keeping functions are both designed and realized.Main work includes invention of prototype vehicle, lidar-based active collision avoidance and GPS-based lane keeping. As for invention of prototype vehicle, system architecture is consist of environment perception layer, high-level control layer and low-level control layer; Low-level control layer is composed of front-wheel control module, rear-wheel control module and HMI module based on modularization and distribution design. These modules are all designed with embedded system; control software is based on uC/OS-II real-time OS, internal communication is based on CANopen protocol. To achieve"dual-mode"switch function, reconstruction of vehicle's mechanical and electrical system is done, including turning system reconstruction,"dual-mode"switch hardware design and power system redesign. The invention process wholly assures reliability,stability and real-time of system operation; openness and expansibility of system architecture.The active collision avoidance part uses lidar as perception sensor; To solve problem of obstacle detection at curve, a servo warning-area detection method is proposed; On prototype vehicle's platform, active collision avoidance is implemented. It adopts two layer decision: impact warning and auto-braking, when comes to emergency situation, computer will take control of vehicle and brakes automatically. Practical experiment shows strong stability of this system.In lane-keeping part, a high-accuracy GPS based lane keeping method is proposed. First, a lane-level digital map is generated using RTK-GPS, road trail information is achieved from map; Then real-time estimation of vehicle's pose is realized by EKF data fusion algorithm. Finally, it judges the degree of lane departure by boundary-crossing time of vehicle's front. When lane departure happens, system will make a warning and take necessary control of vehicle turning system to resume original state in emergency situation. This function is implemented on sample car and practical experiment illustrates strong effectiveness of lane-keeping.