Traffic Modeling For Vehicle Intelligence

Unlike traditional vehicles, research and development on technologies of vehicleintelligence heavily involve vehicle driving environment, especially traffic environment. Itis well known that vehicle surrounding traffic has great impact on the safety andperformance of intelligent vehicles, including object sensing and detection, trajectoryplanning and automatic control. Road test and field experiment remain to be one of the mostwidely adopted approaches in today’s testing and validation on technologies of vehicleintelligence. However, the drawbacks or even the challenges are the high cost, longdevelopment cycle, lack of repeatability, and especially the safety risk it may face. Althoughmodeling and simulation technology has been the mainstream for vehicle development, thetraditional ways may not be adequate and suitable to modeling of vehicle intelligence due tothe fact that they are mainly focused on vehicle performance by itself without muchattention paid on the driving environment, such as surrounding traffic. Therefore, studiesand development a modeling method on traffic for the purpose of vehicle intelligencebecome critical important.Unlike traditional vehicle simulation based on many-body dynamics, and also differentfrom the traditional traffic simulation based on hydromechanics, the modeling method ontraffic for the purpose of vehicle intelligence is just starting. The purpose of this modeling isto provide a credible, real-time traffic environment for the research on object sensing anddetection, trajectory planning and automatic control. Especially, the traffic environmentneed react to the tested intelligent vehicle, so it is full of challenges.With the development of environment modeling technology and sensor technology, thetraffic modeling is put forward new requirements. The traffic not only needed to haverandomness as real traffic flow, but also should have the characteristics of real vehicle dynamics and can control itself actively, etc. What’s more, to be credible and real-time arealso the key words of traffic environment simulation.This paper deeply analysis current status of traffic modeling. And then, research on thetraffic for the purpose of vehicle intelligence with the need of vehicle intelligencesimulation. Based on the research achievement, this paper proposed the modeling methodon traffic which can produce traffic flows meet different macroscopic characteristics basedon the requirement of researcher and intelligent traffic vehicles which can drive and makedecision by self. What’s more, all traffic vehicles can react to both others and the testvehicle, so the experiment environment can be complex and actual.The main work and achievement of this thesis are as follows:(1) Research on the modeling method on traffic and traffic vehicle, and then proposed theirmodels. The proposed traffic model includes vehicle dispatch model and intelligenttraffic vehicle model. The vehicle dispatch model can produce the traffic flow meet theexperiment requirements. And, the intelligent traffic vehicle can drive by itselfaccording to the information from road and other traffic vehicles.(2) Research on car-following model, based on the old IDM model, through changing theacceleration adjustment, solving the lack of old IDM model when vehicle restart.(3) This thesis studies the decision-making method of intelligent traffic vehicle, andestablishes a novel model based on fuzzy method. Proposed the accumulateddissatisfaction of driver as the trigger of decision-making. This model considering bothsubjective and objective factors that affect lane-changing desire of driver. Thesubjective factor is the benefit and the objective one is the character of driver. Use theprobability of collision to judge the possibility of lane-changing. And then, connect thewhole factors by fuzzy inference. So, the decision-making method can simulate thedecision-making process of real driver fully, to make the behavior of intelligent vehiclehave a high degree of confidence. So the model can show the dynamic uncertainty ofreal traffic flow, to make the traffic simulation has a high degree of confidence. (4) For lane-change of intelligent traffic vehicle, this thesis proposed a novel method thatdynamically plans trajectories considering multi-targets. This method generates thetrajectory by a two-degree vehicle dynamics model，so the trajectory can match thevehicle dynamic characters and generated efficiently. What’s more, the trajectory can beadjusted based on different need for the multi-targets such as safety and efficiency. So,with the estimation of other traffic vehicles trajectories, this method can plan an optimaltrajectory even in a complex driving environment quickly. This method not onlyensures quick reaction to the changing driving environment, but also makes the trafficvehicle seem real.(5) Based on the research on modeling method on traffic, this thesis designs and programsto realize the traffic simulation system. Designed the road structure support for systemand the method based on road structure to manage whole vehicles. Final, verified thetheory model and traffic simulation system by testing.