Research On Localization Method Of Unmanned Vehicle Based On Multi-Source Information Fusion

As unmanned vehicles become more and more common in life,the localization of unmanned vehicles is becoming more and more important as a basic pr emise for planning and control.With a single sensor,it is difficult to meet the requirements of complex environments.Multi-sensor fusion is an inevitable trend.In this paper,focusing on the localization of unmanned vehicles in complex environments,which consists of environmental switching,sensor failure,as well as asynchronous and delay,a multi-sensor fusion localization algorithm is constructed based on the factor graph model and incremental smoothing.The positioning accuracy,Robustness and real-time of the algorithm is verified by simulation testing.The specific research content can be summarized as follows:Firstly,the probability model and factor graph model of sensor fusion localization problem is established.The description of the localization problem is given based on the factor graph and the joint probability density.Then the factor representation of the prior information,IMU,GNSS and odometry is analyzed.And from the mathematical point of view,the localization problem is transformed into a nonlinear least squares problem,and the smoothing method is used to solve the process.Secondly,the incremental smoothing algorithm based on the Bayes tree is studied.The factor graph is transformed into a Bayesian network through variable elimination,and then the Bayes tree is established through clique finding.The tree structure of the Bayes tree facilitates probabilistic reasoning,so the processing of new measured values could be discussed.Only the part affected by the new measurement value is updated,so that incremental smoothing and real-time solution is achieved.Thirdly,a multi-sensor fusion framework for unmanned vehicle localization is constructed.For different sensors,IMU pre-integration,visual inertial odometry,lidar odometry are used as preprocessing modules;On accout of possible sensor failure in complex environments,an identification method is proposed.Besides,asynchronous fusion method is also proposed to deal with the different frequency and time delay.Then,the framework of the entire sensor fusion problem is constructed.Finally,the multi-sensor fusion localization algorithm is tested by simulations on some datasets.The positioning accuracy test of different sensor combinations is carried out,and qualitative and quantitative analysis is given.The results show that the fusion algorithm can get higher accuracy than VIO or LO,and the average error is about 20 cm,which meets the index requirements.For localization robustness,we simulate the switching of indoor and outdoor environment,and the results indicate that it can still maintain high accuracy on the XY plane.Finally,the real-time performance of the algorithm is qualitatively analyzed.The factor graph and incremental smoothing enable the fusion algorithm to provide localization results with the frequency no less than 50 HZ.