Underwater Simultaneous Localization And Mapping Of UUV Using Acoustic Vision Information

Accurate positioning information is the basis for supporting the underwater operations of the Unmanned Underwater Vehicle(UUV).At present,UUV uses inertial navigation or dead-reckoning navigation to realize underwater autonomous positioning,but the positioning error accumulates over time.In order to improve the accuracy of underwater autonomous positioning,UUV can correct the positioning error by using the satellite navigation system on a regular floating basis or correct the positioning error in the underwater environment by means of pre-arranged acoustic beacons,or based on marine geophysical navigation,the positional error divergence could be suppressed within the scope of a prior sea chart of geophysical information.Obviously,in unfamiliar sea areas,these navigation technologies cannot meet the demands of underwater positioning in the large-scale environment and long-time voyage.Simultaneous Localization and Mapping(SLAM)technology,by deeply integrating the information of environment-aware sensor equipped by UUV and the information of autonomous positioning,enables UUV to correct the underwater positioning error while constructing map for the environment,and could get rid of the restriction of external navigation beacon or the prior sea chart on navigation ability for UUV,and theoretically,UUV can realize autonomous positioning in unknown underwater environment under long-time and large-scale conditions.In this paper,the underwater SLAM method of UUV is studied based on the acoustic vision information generated by mechanical scanning imaging sonar and multi-beam sonar,and the following sections are studied:Firstly,aiming at the structural underwater environments such as ports and reservoirs,which are common working environment for UUV,using Hough transform to extract line feature from acoustic vision information of mechanical scanning imaging sonar means a great deal of calculation and low efficiency,so a line feature extraction algorithm based on RANSAC-Hough transform is proposed in this paper.Secondly,aiming at the local characteristics of environment perception for UUV and the global requirements of environment map construction,an adaptive method to construct grid map is proposed.The dynamic decomposition technique for space is used to enhance UUV's ability to construct maps of underwater environment.By fusing sub-maps,the grid map can adapt to the scale of the environment.Thirdly,aiming at the inconsistency between map and the dataset of sensor observation in the underwater SLAM algorithm,constrained by the observation ability of acoustic vision sensors,a heuristic ICP algorithm is proposed,which enables UUV to have the ability of self-segmentation for matching point cloud data.By constructing more consistent point cloud data,the matching accuracy and computation efficiency of ICP algorithm are improved.Finally,aiming that the uncertainty of navigation sensor and acoustic vision information is large,and the nonlinear filtering model of underwater SLAM algorithm is prone to mismatch and other problems,the underwater SLAM algorithm framework based on strong tracking UKF is constructed to improve the robustness of SLAM algorithm.And based on the framework of strong tracking UKF,the underwater SLAM algorithms based on feature map,pose and grid map are realized respectively.