Research On Multi-source Information Fusion Navigation Method Of Unmanned Vehicle For Complex Urban Environment

In recent years,self-driving cars gradually come into public notice,self-driving technique has become one of the most popular fields.The complexity of urban environment has an adverse impact on the quality of sensors,and impose challenges and threats on high-precise navigation and positioning of self-driving cars.At present,the type of sensors and method of sensor fusion used in cars are both single,which cannot accommodate self-driving cars to navigation robustly and precisely in complex environment,thereby it is necessary to introduce and make the most of various redundant information from types of sensors in the multisensory fusion systems,which provides more comprehensive and valuable navigation solution.This paper is aimed at multisensory fusion used in self-driving cars in complex urban areas,conducts research on multisensory error modeling,GPS time-differenced carrier phase high-precision positioning,multisensory fusion framework and algorithms,so as to enhance the reliability and precision of the navigation system of self-driving cars.The research on error characteristics of car-mounted sensors is the basis of multisensory fusion.The navigation system in self-driving cars may be interfered with the complex conditions in urban areas,such as sideways buildings,pedestrians and obstacles,it is hard to build error model which accommodates to environment,thereby the research on error modelling and compensation of various sensors in special environment is conducted at first.The sensors applied into vehicles mainly includes low-precise MEMS-inertial sensors,GPS,odometer,camera and wireless positioning system.Therefore,the selection and error analysis of sensors are completed at first,the intrinsic and extrinsic errors are diminished,so as to ensure the accuracy of multisensory fusion.The research on error characteristics of car-mounted sensors is the basis of multisensory fusion.The navigation system in self-driving cars may be interfered with the complex conditions in urban areas,such as sideways buildings,variety and complexity of roadways,the accuracy of navigation systems in self-driving cars are heavily affected in actual conditions.Therefore,based on commonly used navigation sensors,including MEMS-INS,GPS,odometer,camera and UWB,this paper carries out research on error modelling and compensation of various sensors in special environment.One the one hand,the selection of sensors is designed based on error characteristics of different kinds of sensors.On the other hand,the error calibration and compensation methods are studied and the intrinsic and extrinsic errors are diminished,so as to ensure the accuracy of multisensory fusion.GPS is one of the most common positioning methods in cars,however,it is easily influenced by occlusion and interference,and its accuracy is limited in single mode.Considering this fact,this paper conducts the research on high-precision GPS positioning method based on carrier phase.First,according to the deriving model of carrier phase measurements,carrier phase is differenced between the consecutive epochs to mitigate the integer ambiguity to obtain high-precision positioning results.Then,its result is regarded as measurements for GPS/INS integrated navigation system.In addition,aimed at the cycle slip problems of carrier phase easily occurred in occlusive environment,an adaptive cycle slip detection and repair method aimed to solve vehicle dynamic characteristics is proposed.It can reduce the false detection rate of cycle slip and ensure the quality of observations for back-end fusion.On the basis of initial error calibration and compensation of various sensors,the research on multisensory fusion scheme using factor graph is conducted.At first,the graph-optimization based algorithm is studied,the factor nodes representing sensors are established to compose a multisensory fusion graph model,the full optimization for all states is implemented and the errors of sensors are estimated,calibrated and compensated simultaneously.Second,aimed at problem of unsatisfactory current state estimation,this paper proposes a modified scheme in which IMU measurements are used to recurse the current states on the basis of past smoothed states.Finally,the simulation and experiments are conducted to prove the proposed method.In order to verify the proposed multisensory fusion method,on the basis of the above schemes and algorithms,the car-mounted testbed is built and field tests are carried out.The results show that the proposed method can effectively enhance the positioning accuracy of vehicles and ensure the reliability and precision of self-driving cars in urban areas.This paper is aimed at the multisensory fusion scheme and method of self-driving cars in urban complex areas and the theory is proven by the experiments.It has a good prospect on engineering application.