Research On Intelligent Vehicle Localization Method Based On Multi-sensor Information Fusion

The vehicle positioning module is the foundation of the intelligent driving car system and one of the most important components.It is responsible for solving the problem of vehicle perception of its own position and posture,and provides important information for decision-making,planning and control modules.At present,intelligent driving vehicles are equipped with various types of sensors to capture environmental information,and vehicle positioning is achieved through information fusion methods.The type of sensor and the method of information fusion directly affect the positioning accuracy of the vehicle.Therefore,the development of vehicle positioning technology research based on multi-sensor information fusion methods to improve the positioning accuracy of vehicles in complex traffic environments will help promote the process of large-scale commercial applications of intelligent driving vehicles.In order to improve the accuracy of vehicle positioning on urban roads,while reducing the uncertainty of position estimation,this paper conducts in-depth research on intelligent vehicle positioning methods based on the idea of multi-sensor information fusion.The main research contents are as follows:(1)A localization method based on cooperative Bayesian filter is established.The principle of Bayesian filtering algorithm is introduced,and on this basis,a cooperative Bayesian filter is proposed that combines original GPS data of the main vehicle,position data of surrounding vehicles,and lateral distance data from the vehicle to the lane line.At the same time,an Open Street Map-Pre Scan-Simulink co-simulation platform was built.With the help of Open Street Map map data,a predefined fixed traffic flow and a random traffic flow based on Gipps car following model were established.The simulation results show that the cooperative Bayesian filtering can greatly reduce the position estimation error of the original GPS,and this method can be used as a preprocessing step for other positioning methods(2)A localization method based on Augmented Extended Kalman Filter(AEKF)is proposed.Four common-used motion models in Kalman filtering are introduced:Constant Velocity(CV),Constant Acceleration(CA),Constant Turn Rate and Velocity(CTRV)and Constant Turn rate and acceleration(CTRA).Aiming at the problem that a single motion model cannot be applied to all driving scenes,a multi-motion model fusion algorithm based on T-S fuzzy inference mechanism is proposed.The algorithm takes the acceleration and yaw rate of the vehicle as input and is fuzzified by Gaussian membership function.After the post and center of gravity defuzzification method,the predicted value of the current vehicle system state is output.On this basis,combined with the predicted value of the fusion of multiple motion models and the measured value of the cooperative Bayesian filter,an Augmented Extended Kalman filter is proposed.Experimental results show that the AEKF positioning method can reduce the original GPS position estimation error from5 m to sub-meter level,which meets the positioning requirements of smart vehicles.(3)In-depth analysis of the cooperative Bayesian filtering and enhanced extended Kalman filtering proposed in this research: firstly,the influence of the number of lanes,the number of surrounding vehicles,and the number of particles used in solving the cooperative Bayesian filtering on the positioning accuracy and running time of the algorithm is analyzed.Then it discusses the positioning performance of the combination of cooperative Bayesian filtering and other positioning algorithms,and finally proposes a method to improve the cooperative Bayesian filter by using lane data.