Research On Simultaneous Localization And Mapping Method Of Autonomous Underwater Vehicle

Under the increasingly competitive international situation,underwater navigation has a very high research value and significance for civil salvage and rescue,submarine pipeline maintenance,and military submarine strategic cruise.Now it has become a hot and difficult research field at home and abroad.According to the characteristic that the inertial navigation system can meet the requirement of positioning for a long time,the combination of the SINS(strapdown inertial navigation system)and SLAM(simultaneous localization and mapping)is investigated in this paper.The SLAM technology can not only provide the position information of carrier and road sign,and assist in correcting SINS drift error,but also can be used to construct the characteristic map which is consistent with the surrounding environment.Based on the research background and significance of the subject,this paper summarizes the data association algorithm of SLAM at home and abroad,and has carried on the thorough research to its key technology.According to the background and significance of detecting and locating engineering for submarine oil pipeline,a pipeline model with certain functions is set up and the function trajectory is taken as the navigation reference path of the carrier.The azimuth information,distance information and image information are collected by acoustic,optical and inertial sensors,respectively.Simultaneously,the multi-sensor fusion algorithm is researched,and the simulation experiment realizes high-precision pipeline positioning.The specific improvement algorithms and research contents are as follows:1?In the underwater navigation and positioning,the carrier must carry a variety of sensors to complete collaborative navigation tasks.However,among the observed values of the location of these sensors provide,the data association analysis is need to determine whether they are from the same target,and then the observed values are stored in the map database.Aiming at the shortcomings of nearest neighbor association algorithm in dealing with numerous beacons and in the complexity of association,a novel SLAM data association algorithm based on raster graph fuzzy logic is proposed.The simulation results show that the algorithm can achieve high precision data association.2?With regard to the ambiguity of underwater environment and system observation noise,it is easy to generate fuzzy association or error when data association is done and then to lead to the failure of SLAM process.A data association algorithm based on maximum probability BP neural network is proposed.The experimental results show that the new algorithm can obtain a high correlation degree when the observation values of target position are associated,and its corresponding position estimation filter precision is also quite high.3?As there are the problems that the non-linearity error is large,the time updating and prediction updating are cumbersome and the navigation computer is slow for the filter methods of EKF-SLAM,UKF-SLAM and ?-UKF-SLAM,a particle filter-based SLAM algorithm,which utilizes a minimum set of sample points to approximate the distribution function of the system state,is put forward.In order to satisfy the particle diversity and abundance,it is judged whether the resampling is carried out during the updating of particle weights by setting valid samples.In view of the application of underwater pipelines detection and positioning,the simulation experiments prove that this algorithm can teach a high positioning accuracy in diagonal function pipeline position.4?For the fuzzy uncertainty of the system itself of large area of the seabed or the random signal model,the estimation accuracy of initial state is low and the robustness of the system is poor.Combined with the pipeline detection and positioning of the "curve" type function,a SLAM algorithm based on an extended H? filtering is proposed.The algorithm designs the H? filter of the minimum norm from the interference input to the filter output,and performs time update and forecast update by obtaining the corresponding system parameters.The simulation results show that the trace detection has high correlation precision and the partial pipeline characteristics maps collected and extracted can reach high accuracy.5?For the problem that the dimensionality of sampling points in nonlinear filtering of traditional n-dimensional sigma point's is large,it leads to increaseing calculation work and decreasing the computation speed and results to that the observability of the system under large noise is weak.A new square-root Cubature Kalman filter(SR-CKF)method based on Spherical Simplex(SS)of SLAM is proposed.This algorithm improves the run speed of the navigation parameters of the model by computing and propagating the cubature points of n+2 dimension,predicting the state and updating the estimation,and improves the speed of the navigation parameters in the model.Compared with the SRUKF and SSCKF filtering algorithms,the experimental results show that the proposed SS-SRCKF algorithm has higher accuracy and better stability under the square-root covariance guarantee of non-negative dominance,and it also proves that the new algorithm is effective in the future application of underwater navigation engineering.6?According to the research background of underwater pipeline detection,it sets up a specific simulation environment,and three kinds of filtering algorithms,particle filter,extended H? filter and SS-SRCKF are used to estimate the position of the carrier respectively in the simulated region of the pipeline path.Based on the idea of feature extraction in scan invariant feature transformation algorithm,the underwater environment is simulated,and the characteristics of the aircraft wreckage and pipeline in the image are extracted,and the feature map in SLAM is constructed.