Researches On Driver Behavior Modeling

The traffic accident take more than 80% of the total accidents, it becomes "the top killer" ofhuman being. The research indicated that more than 95% of traffic crashes result from driver'smistakes. It proved that the researches on driver behavior take active effects for traffic safety.People have taken focus on driver behavior researches for a long time. The progress has beenmade in the traditional area of drive behavior researches. Some achievements have beenapplied to auto manufacture industry. With the outburst and rapid development of computerscience and life science, the researches on application of cognition science in the auto industryhave become one of the hottest research spot. One of the most important purposes ofresearches on drives behavior is to resolve traffic safety problems. An independent research ondriver behavior is meaningless at all. The model's integration and its application are the goalsand trends for driver behavior's researches. This paper details the following six aspects ofdriver behavior researches:First, comfortable and safe driver behavior model of car-followingCar-following is the most common driver behavior, and it's a typical research area of driverbehavior, and the car-following has become the bases of many research realms. This part ofthe paper begins with topic of comfort and safety on which the industry emphasized andinterested, and starts the researches on physical safety and physiological discomfort caused byhard brakes of car-following. On the base of kinetics laws and empirical data, and the boundaryconditions being considered, the procedure of safe and comfortable driver behavior modeling isdetailed. The experimental methods are used to verify the model's validation. The concept ofpervasive computing is also introduced to driver behavior modeling to elaborate new ideas inthe auto applications. The purpose of this part of research is to point out that the traditionalresearch methods is the important parts of driver behavior researches, and its researchachievements should be referenced by driver behavior modeling in cognitive architecture andintegrated in the fusion of driver behavior models.Second, urgent driver behavior modeling in cognitive architectureThe urgent circumstance seldom happens in driver's career. However, since the most crashesoccur under urgent circumstance, the researches on driver behavior modeling under urgentcircumstance are especially important. ACT-R is a typical cognitive architecture, and beingwidely used for its so many advantages. This part of the paper elaborates the process of driverbehavior modeling under urgent circumstance based on ACT-R cognitive theory, and both thedriver behavior's prediction over model and model's verification with experimental methodproved that the model is widely suitable and flexible. The illustration of the benefits of drivebehavior modeling in cognitive architecture make it an assertion that the application of driverbehavior model in the field of traffic safety will be set up effectively and take effects greatly.Third, Multi-tasking driver behavior modeling in cognitive architectureBy analyzing the driver's perception, cognition and decision-making, and sensorimotor, itelaborated the fact that the driver task of human system is handled in parallel. By using the methods of classification of driver task, rank of risk level, and introduction subconscious toACT-R model, and ACT-R is improved with multi-production-system, multi-monitor module, andmulti-motor module to realize parallel process of multitasking driver behavior. Not only theexecutive efficiency is improved, but also it is of coincidence with the real world. The method ofJava multi-threading is applied to deal with parallel multitasking, and the software simulationmethod is applied to verify the validation of the improved model.Forth, knowledge discover of driver behavior in traffic accident databaseThe knowledge discover of driver behavior in traffic accident database is implemented basedon Multi-Agent technology and KDD~* model. To solve the complexity problems of both discoverand driver behavior, the Multi-Agent technology is applied. This method helps knowledgediscover of driver behavior to realize intelligentize. The application of KDD~* model makes theknowledge discover go forward in the direction of the system's automation convergency and itcan improves the whole efficiency of discover process. Multit-hread technology, different ofKDD methods are introduced to knowledge discover of driver behavior in traffic accidentdatabase. The results show that KDD in the database of traffic accident based the abovemethods is effective. The beneficial knowledge to researches of driver behavior is obtained.Fifth, software simulation of driver behavior modelingSimulation is a powerful tool for evaluation and analysis of new system designs, modificationsto existing systems and proposed changes to control systems and operating rules. The aims ofthis part of the paper is to build simulation and imitation environment of driver behaviorresearches, and to make research works more intuition and convenient. On the basis ofcomparing and analyzing of exist cognitive modeling tools, it emphasized the importance forthe development of cognitive simulation software. Then the paper describes the implementationprocess of the simulation and imitation environment from the system's construction viewpoint.The environment mainly includes the following functional modules at present, behaviormodeling, model's modification, model's verification and scenario construction module etc. Themain supported theory of the environment is ACT-R cognitive architecture.And finally, fusion of driver behavior model and its applicationIntegrating and applying driver behavior model, and improving the transportation capability anddecreasing the traffic accident are the final goals of driver behavior researches. In this part ofthe paper, we want to build such a kind of platform, and design a practice course to improve thedriver's performance over such a platform. Firstly, the paper elaborate the platform's idea,which driver behavior researches, driver behavior modeling in cognitive architecture,experimental verification, software simulation, and all their application are integrated to aplatform to lead the research purpose to his original which solve the traffic accident problems.And then, the whole practice course is detailed and practice results are compared. It provedthat what the most important of driver behavior research is driver behavior models' integration.