Three Dimensional Positioning And Tracking Of Vehicles For Intelligent Driving

With the development of electronic information technology,the intelligent driving has received extensive attention nowadays.Intelligent vehicles have rich sensing capabilities that can interact with other vehicles and ground infrastructure,etc.Based on the vehicle information network,autonomous vehicles can not only reduce the occurrence of traffic accidents,but also improve the traffic and energy efficiency.However,the implementation of fully intelligent driving still faces with many key problems such as the high-precision three-dimensional spatial position and the contour information of the vehicle.In complex urban environments,satellite navigation is unable to meet practical needs due to the signal attenuation by buildings and other obstacles.In addition,as a real object with non-negligible length,width and height,the vehicle can no longer be viewed as a point target but an extended target.It is necessary to analyze the acquisition methods of contour information and impacts in practical applications.High-precision three-dimensional spatial position and vehicles' contour are necessary for intelligent driving.Therefore,based on the Bayesian framework,this dissertation proposes three dimensional positioning and tracking of vehicles for intelligent driving from the perspective of probability,which effectively enhances the three-dimensional position precision and the contour fitting of the vehicles.In view of the three-dimensional positioning and tracking issues,the “vertical”measurements from unmanned aerial vehicle(UAV)are adopted for assistance.Based on the Belief Propagation(BP)framework,the algorithm takes advantage of the distance obtained by the Time of Arrival(TOA)and the relative position of the UAV,by which the 3D position of the vehicle can be achieved.The 3D UAV-aided vehicle positioning and tracking based on UAV algorithm is firstly verified by the Cramer-Rao Lower Bound(CRLB)analysis.Furthermore,the ways of message passing between vehicles and UAVs are discussed.On the conditions of modern 5G mobile communication network,we try to study different ways of messaging respectively,so that the algorithm can achieve the best drone-assisted strategy and meet the high precision requirements of intelligent driving.On the other hand,aiming at the requirements of the 3D extended vehicle,this dissertation proposes a 3D extended target tracking algorithm based on random matrix framework.The algorithm tries to fit the vehicle scatters using an adaptive ellipsoid,so that the complex problem of extended target tracking is decomposed into the motion problem of particles and the estimation problem of the contour.Then under the prediction and update of the Bayesian framework,We try to realize the position update and contour fitting of the vehicle effectively in real time.As the result,the precision of the vehicle's contour reaches decimeter level,which is suitable for the tracking scenesfor intelligent vehicles.This dissertation mainly focuses on the three-dimensional position and contour information of intelligent driving in complex environments.Through the main investigations,the algorithm can deal with the 3D positioning and tracking problem of vehicles in complex environment,and provide theoretical supports for the intelligent control and optimization framework in the future.