Research On Cooperative Collision Warning System Of Vehicles On Highway

The independent collision warning system that based on radar-ranging is hard tobe universal for its high cost and the limitation of line-of-sight range and environment,also it is commonly considered to be ineffective in the chain rear-end collision whichis very familiar on the highway. Meanwhile, the “internet of things” or the network of“vehicle-to-vehicle (V2V)” was proposed and is going ahead fast. On this context, theidea of cooperative collision warning is put out and thought of to be a favorableresolution on collision warning. In this paper, the main function modules areresearched, which include the warning algorithm, the V2V communication base onZigBee technology, the lane identification based on machine vision, the rangingbetween two neighbor vehicles based on the relative GPS position, and the focus isput on the last two one.Firstly, in the research, underlying causes of collision accidents on highway areanalyzed, and then the warning system structure is designed and described briefly.After that, function modules of the system are researched in each chapter. Thedetecting of current lane and vehicle distance is introduced in the3rdand4thchapter.In the next chapter, the ZigBee wireless communication technology and the ZStackprotocol stack are introduced. Vehicles transfer their motion state information andreceive it from their neighbors through ZigBee wireless communication. By this way,the interchange of the motion state between neighbor vehicles is realized. Finally, thewarning algorithm and the safety-distance model is put out, which apply to thiscooperative collision warning system.The line identification is realized based on the machine vision and image process.The camera mounted on front-end of vehicle catch the image of the front road in realtime, and then the line identification unit makes an image processing, which firstlythe line feature is detected as candidates for line marker using Hough Transformmethod, then the candidates is verified by the geometric features of line markers.Finally, the effectiveness of the remained lines for lane identification is verified, andthen by using the distribution of slope’s values, the current lane was identified.In GPS position, two single frequency C/A code civilian GPS receiver is used.The navigation information is obtained by MCU via serial communication. Before thedetection of vehicle distance using GPS relative position, the performances of GPSstatic position and dynamic position and the interrelationship between different GPSmodule are tested. Then the process of relative motion of two vehicles is simulated. The data got in this process is collected and the property of GPS relative position isanalyzed. The result suggests that GPS relative position have the potential for rangingbetween two vehicles. Finally, the Kalman filter which is commonly used for filteringand data fusing in the navigation and target tracking is introduced, togethermaneuvering target movement model and GPS position error model are introduced.In the end, the work in this paper is summarized. The inadequacy andimprovement of it are pointed out, and the application of cooperative collisionwarning technology is anticipated.