Research On The Key Technology Of Cooperative Localization For Intelligent And Connected Vehicle In Urban Scenarios

Connected and automated vehicle(CAV)are equipped with advanced vehicle-mounted sensors,controllers,actuators,and other devices which integrate modern communication and network technologies to realize intelligent information exchange and sharing between vehicles and X(pedestrian,vehicles,roads,and clouds,etc.).Environment’s perception,intelligent decision-making,cooperative control,and other functions can realize safe,efficient,comfortable and energy-saving driving,and finally realize a new transportation to replace drivers.Comprehensively improving vehicle safety,economy,comfort,and road traffic efficiency has posed a new challenge for the development of the CAV industry.However,perceptual localization technology has developed into the pioneer of the CAV industry.To achieve accurate perception,efficient decision-making and stable control of vehicles,the networked cooperative technology of vehicles plays an important role.The localization accuracy of vehicles in urban scenarios is easily affected by driving environments such as dense forest coverage,high-rise buildings,overpasses,tunnels,and underground parking lots,so that the vehicle cannot accurately perceive the location information of the vehicle.Therefore,there are still many scientific problems to be solved for vehicle accurate localization in complex urban driving scenarios.Dynamic and static vehicle localization under multi-dimensional transformation is primarily difficult to form a unified architecture for universal adaptive coordinated distance precise positioning.In time-varying driving scenarios,including tunnels,underground parking,and other signals interrupted for a short period of time,the vehicle cannot obtain real-time position information.The received signal of the urban high buildings is affected by the multi-path effect,and the position information of the vehicle has large errors.Aiming at the above positioning-related problems,this paper takes multi-vehicle,multi-base station and other systems as the research object,and studies the acquisition of accurate real-time vehicle position in the case of multi-source information fusion such as multi-vehicle and multi-base station.It mainly starts from different driving scenarios such as maintaining the safe distance between vehicles,realizing accurate adaptive switching of multi-road scenarios,considering the effect of the vehicle speed and the number of base stations on the vehicle localization accuracy.It provides a theoretical basis and technical support for the accurate positioning and perception of intelligent networked vehicles.The main contributions of this paper are summarized as following:(1)The vehicles cannot accurately obtain the distance of inter-vehicles,in-depth study of the localization mechanism inter-vehicles,and the position interaction mechanism between vehicles is studied.The information coupling relationship between vehicles,and the base station is explored.Based on the direct difference method,single-difference pseudorange measurement method,double-difference pseudorange measurement method,and the theory of pseudorange difference measurement method,a unified cooperative localization and ranging architecture is designed.The designed architecture can realize the adaptive switching of methods under different driving conditions,form a cooperative localization system of inter-vehicles distance,and realize the precise measurement of vehicle-to-vehicle distance based on cooperative localization.(2)The traditional Kalman filter algorithm is difficult to achieve accurate positioning in the process of vehicle motion,an adaptive interactive multi-model Kalman filter and multi-base station arrival direction fusion estimation algorithm is proposed.Based on the unbiased estimator,the measurement noise covariance is updated in real time and embedded in the standard Kalman filter algorithm to realize the application of adaptive interactive multi-model Kalman filter.Considering the influence of different vehicle motion states and dynamic driving environment on the accuracy of vehicle positioning estimation,an adaptive interactive multimodel Kalman filter and a multi-base station information fusion algorithm are constructed to estimate the vehicle position with different vehicle velocity and numbers of the base station in real time.(3)The urban extreme driving scenarios(urban forests,tunnels,overpasses,underground parking lots,etc.),the short-term signal interruption of the vehicle leads to the large error of vehicle location recognition.The vehicle position is estimated based on the mathematical geometric topological coupling relationship such as vehicle kinematics and road shape,which reduces the position constraints on environment perception in the process of co-localization.A relocation method is proposed to increase the positioning accuracy under GNSS signals.The effectiveness and scalability of the co-localization algorithm under different driving conditions are evaluated.In the case of short and long GNSS signal loss,the cooperative localization algorithm can effectively handle special GNSS signal loss driving scenarios.(4)The received signal of urban high-rise buildings is affected by multi-path effect,non-lineof-sight,etc.,which makes the vehicle obtain the position of the vehicle with large errors,a semi-definite relaxed-received signal strength convex optimization method is proposed.Based on the distance measurement of the received signal strength,the multi-source least squares problem is reconstructed,and a new convex constraint is designed to strengthen the non-convex maximum likelihood estimation algorithm to realize the convex relaxation solution.Meanwhile,the influence of completely unknown and completely known non-line-of-sight state information on the positioning accuracy is considered.An error detection method is proposed in the process of detecting the non-line-of-sight path of the target vehicle.This method has remarkable stability in reducing non-line-of-sight errors.