Research On Key Technologies Of GNSS/INS/LiDAR Multi-source Integrated Navigation System

The multi-source integrated navigation system integrates multiple navigation sources,which realizes the sharing of information and complementary of performance between internal sub-navigation systems,and can provide continuous and accurate navigation solutions for vehicles,airborne,missile and other platforms in complex navigation environments.The multi-source integrated navigation system has become an important means of autonomous navigation and has been widely used in both military and civilian fields.Based on the actual needs of airborne navigation systems,this paper aims to improve the accuracy and robustness of airborne navigation systems with the framework as multi-source integrated system.In the paper,key technologys as the Light Detection and Ranging system,the Li DAR/INS integrated navigation system,the information fusion algorithm and the fault detection algorithm are studied.The main innovations of this paper include 4 aspects.(1)Aiming at the problem that INS output attitude error affects the accuracy of navigation estimation precision in the existing feature points based Li DAR airborne navigation algorithm,an altitude-distance matching algorithm based on Scale-invariant feature transform(SIFT)is proposed,which directly matches the interpolated Li DAR measurement range data with the locally stored Digital Elevation Map(DEM)data.In the algorithm,the fuzzy control is used to calculate the threshold of the SIFT algorithm,and the optimal matching point cloud is extracted by multiple cycles.On this basis,to further improve the precision of the Li DAR navigation system,a positioning algorithm based on robust ridge estimation is proposed.The proposed algorithm transforms unbiased estimation into biased estimation by introducing ridge parameters,which realizes optimal position estimation under the constraint of minimum root mean square error,and reduces the effect of poor quality observations on positioning accuracy by using iterative weights.Simulation results show that compared with the existing Li DAR airborne navigation algorithms based on feature points,the proposed two algorithms can effectively reduce the dependence of the Li DAR system on the INS system,and further improve the positioning success rate and positioning accuracy of the airborne Li DAR navigation system.(2)Aiming at the problem that the performance of feature-based Li DAR airborne navigation algorithm is affected by platform inclination and ground surface undulation,this paper proposes an adaptive switching algorithm based on Sigmoid function,which constructs an adaptive switching model by judging the carrier attitude and the surface area undulation information of the scanning area,thus improves the stability of the Li DAR/INS integrated navigation system.In addition,based on the above research,two feasible Li DAR/INS integrated navigation methods are analyzed in detail,and their performances are verified and compared through simulation experiments.The simulation results show that the two methods have their own advantages and disadvantages in different navigation scenarios.(3)Aiming at the problem that the existing multi-source integrated navigation based on factor graph is affected by the time-varying subsystem measurements noise,which resulting in a sharp decline in the accuracy of state estimation when the navigation environment changes,this paper presents a method for estimating the mean vector and covariance matrix of subsystems based on Gaussian model in the framework of factor graph.In the proposed method,the maximum posteriori estimation of the mean vector and the covariance matrix of the subsystems are updated in real time by using the navigation residuals during each iteration period in the process of factor graph optimization,so as to obtain more accurate navigation state estimation.Both the simulation test and the experimental test results show that compared with the existing factor graph algorithm,the verified factor graph based on iterative maximum a posteriori estimation can effectively improve the multi-source integrated navigation estimation accuracy when the subsystem measurments changes along with time.(4)Aiming at the problems of low real-time detection and poor effect of gradual fault detection in existing neural network-based fault detection algorithms for multi-source integrated navigation systems,a multi-channel Single-dimensional Fully Convolutional Neural Network(MS-FCN)fault detection mentod is proposed.The proposed method takes the residual sequence in the integrated navigation system as input,uses different size of convolutional neural networks to extract feature information with different scales including local and global in the residual sequence,and restores the size of feature map as same as the input by deconvolution.Thus,the proposed method enables a comprehensive and accurate diagnosis of the operating state for each sampling point.Simulation and experimental results show that the proposed MS-FCN has better detection rate and false detection rate for both gradual and abrupt faults comparing with the existing algorithm,thus,the multi-source integrated navigation system based on MS-FCN will have higher stability and navigation state estimation accuracy.