Study On Forward Collision Avoidance System Considering Drivers’ Collision Avoidance Behavior Characteristic

Traffic safety has always been the most serious problem in current society, statisticsshow that70%to90%of accidents are caused by pilot operational errors. Forward collisionavoidance system is an advanced vehicle safety driving assistance technology, which canprovide warning to the driver when making a high likelihood of collision judge, and promotethe driver’s braking operation, even if the driver does not brake, it will still provideautomatic braking to reduce the speed of the vehicle during the impact. Therefore, the risk ofrear-end collision can be effectively avoided, and traffic safety will be improved. Due to itssignificant role in improving automotive active safety, research and development ondecision-making and control algorithm of the system with good adaptability to the drivershave become hot issues in automotive engineering field.According to the present development status of forward collision avoidance system inthe world and combined with the related theory and technical issues, a set of development offront anti-collision system have been studied in this paper, which include analysis andmodeling of drivers’ braking moment characteristics in virtual car following collisionavoidance scenes, definitions of system function, pre-crash warning and automatic brakingdecision-making and development and simulation verification of automatic braking controlalgorithm, development of control system hardware circuit/software, design of printed circuitboard, trail-manufacture of prototype and field test.Based on results of analysis and modeling of drivers braking moment characteristic, theultimate goal of this research is to design anti-collision warning and automatic brakingfunction strategy for forward collision avoidance system. This paper will establish the timelyadaptive method of individual dynamic characteristic difference, propose braking controlvariable decision algorithm based on gray target analysis, develop a prototype with thesystem of collision alarm and automatic braking function which is equipped with adaptivedriver’s individual difference and carry out the system in real car to verify. The system isable to meet the specific needs of the driver personalization, to reduce the false alarm rate, topromote the development and improvement of early warning systems and automatic brakingfunction, and further to improve the traffic safety and efficiency. Three kinds of accident conditions of embedded with real driver’s rear-end crash tests,including approaching the front stationary vehicle, the low-speed vehicle and following thefront slow down vehicle, were conducted by the using of car driving simulator. Corrrelationamong the driver action time information based on the different requirement of theanti-collision braking operation to take, human factors (age, sex) and envirmental factors(motion state information of leading and following vehicles) were studied. Vehicle safetydriving indicator at the time of maximum braking deceleration moment is proposed, BPneural network models were established according to the indicator at the normal and harddriver braking deceleration moment, which will further contribute to understand the behaviorof drivers anti-collision action characteristics, and provide a new feasible solution for theuncertainty and nonlinear characteristics of human cognition and dangerous judge behaviorof rear-end that are difficult to describe. But also it provided a theoretical basis for thesystem development.In order to make the pre-warning function adapt to the braking moment characteristicsof different drivers, different braking stages, the aim of this stddy is to improve theperformance of driver assistance systems. An approach of normal braking decelerationmoment of BP neural network model’s weight parameters timely adjustment was proposedbased on error back propagation algorithm. In order to ensure that the selected trainingsamples can correctly describe the driver braking moment characteristics, the driver normalbraking moment of identification analysis method was presented, so a number of processeddata was used to constitute the training sample set. Therefore, the developed warningstrategy has better robustness, and thus improves the prediction model’s dynamicadaptability.To solve the possible problem of mutual interference between the automatic brakingfunction and the driver to take emergency steering in the anti-collision process, automaticbraking control algorithm was developed by analyzing the characteristics of the driver’semergency steering behavior, which considered a minimum comfort safe distance. Optimalbraking control amount decision algorithm based on the gray theory is proposed to realizeautomatic braking function, which was referenced to the thinking mode of drivers’identification and decision. Then the optimal weights of algorithm were identified accordingto driver’s collision avoidance behavior characteristic by using particle swarm optimization,so reasonable premise for braking control decisions were provided. Based on the optimaldecision result and combined with the establishment of vehicle reverse braking systemmodel, the desired brake pressure was obtained to provide a guarantee for implementing effective anti-collision braking.Furthermore, to validate the effectiveness of main research achievements and systemfunctions, a real vehicle test platform for the system, based on the structure of a conventionalvehicle, was constructed. Several key technologies are developed to improve the flexibilityand reliability of the platform, which included electro-hydraulic braking actuators andcontrollers, radar information processing method, the way of warning selection and so on.Finally, based on these studies above, simulation and real vehicle tests were carried out.The experimental results show that two-stage warning and automatic braking function isreasonable, and the system has the ability to meet the driver braking behavior characteristics.