Collaborative Localization Method For Intelligent Vehicles By Dynamic V2V Information Sharing

As an important development direction for the reform of the automobile industry,intelligent vehicles are committed to achieving safe,comfortable,energy-saving and efficient driving.As the core key technology of intelligent vehicles,high-precision and high-reliability positioning is an important guarantee and requirement of intelligent driving.In the face of long tunnels,urban canyons lined with tall buildings,and densely tree-lined roads,positioning method with a single satellite navigation cannot meet the positioning needs of intelligent vehicles.Therefore,multi-source information fusion positioning methods have become a research hotspot.In a complex traffic environment,there are certain limitations to relying only on a single vehicle for autonomous positioning.Therefore,in this paper,we propose a collaborative positioning method based on the dynamic interaction of vehicle-vehicle information in a networked environment.The main work of the paper is as follows:Using multi-sensor data fusion,considering the problem of multi-sensor time synchronization,we propose a vehicle autonomous positioning method based on the combination of multi-source information and multi-motion model.Aiming at the different motion states of the vehicle,combined with Constant Velocity,Constant Acceleration,Constant Turn Rate and Velocity and Constant Turn Rate and Acceleration models,we researched the interactive multi-model filtering algorithm based on extended Kalman filter,which effectively integrates GNSS and IMU data and realizes the adaptive positioning of the vehicle under the multi-motion state;on this basis,added with LIDAR data,the error state Extended Kalman filter is introduced to correct the positioning result,which improves the precision of positioning and the accuracy and stability of the system.Using the vehicle-vehicle information interaction in the networked environment,we proposed a multi-vehicle collaborative positioning method based on federal Kalman filter.We theoretically analyzed the positioning models when surrounding vehicles are one or more,that is,the positioning with a single-neighbor vehicle and the collaborative positioning with multivehicle dynamic information interaction,respectively.In the single-neighbor vehicle positioning,we established the relative position measurement model of the neighboring vehicle and realized the estimation of the fusion of vehicle’s relative position and absolute position.We also extended single-neighbor vehicle positioning to multi-neighbor vehicle positioning and proposed a multi-vehicle collaborative positioning filter algorithm with multi-vehicle dynamic interacted information.We fully considered the impact of the following four factors—the number of neighboring vehicles,GNSS accuracy,ranging accuracy and communication delay—on our method,and built a trust factor screening strategy to select high-confidence measurement information to solve the positioning results,so as to achieve positioning optimization in a collaborative environment.For the intelligent vehicle positioning method proposed in this paper,we carried out simulations and real vehicle test verifications.In the simulation part,we built our simulation platform based on Carla.With different experimental sensor parameters such as GNSS,IMU,LIDAR,etc.,we compared and analyzed the vehicle autonomous positioning method and collaborative positioning method in the urban environment.In the real vehicle testing part,the verification test of the autonomous vehicle positioning method was carried out on the "Xinda" intelligent vehicle platform,which is independently developed by our group,on the intelligent vehicle test field of Chang’an University.The simulations and real vehicle tests show that the collaborative positioning method based on the dynamic interaction of vehicle-vehicle information proposed in this paper makes up for the shortcomings of a single-sensor and a single-vehicle autonomous positioning method,and can achieve high-precision,high-reliability positioning in complex environments such as tunnels and cities,which provides a new method for vehicle positioning in a networked environment.