Research On Simultaneous Localization And Mapping Of Unmanned Underwater Vehicle Based On Random Finite Set

Autonomous navigation is one of the key technologies to realize the complete autonomy of Autonomous Underwater Vehicle(AUV).Due to the increasingly complex requirements of marine missions,Traditional navigation methods have been difficult to meet the needs of unmanned underwater vehicle to carry out marine missions.Simultaneous Localization And Mapping(SLAM)has become a research hotspot in the field of unmanned underwater vehicle navigation due to its simple requirements for sensors and the advantage of not requiring prior map information.This paper focuses on the underwater SLAM algorithm based on random finite set.Firstly,in order to solve the problems that the underwater SLAM based on the grid map fails to fully and effectively express the observation information of the forward-looking sonar sensor and does not consider the uncertainty of the number of states,the random finite set theory mostly used for target tracking is applied to the underwater SLAM based on grid map,and the framework of the underwater SLAM algorithm based on random finite set is proposed.The random finite set theory can better deal with missing detection and false alarm imperfect observation information in low-accuracy forward-looking sonars,and can better estimate the number and probability uncertainty of states in SLAM.Secondly,for the problem that the computational complexity increases exponentially with the increase of the number of states in Bayesian filtering based on random finite set,The probability hypothesis density is used to approximate the state posterior probability density,and the particle filter is used to implement the probability hypothesis density filtering to reduce the complexity of the algorithm.At the same time,the proposed distribution of the underwater SLAM algorithm based on random finite set is improved,and a maximum posterior finite resampling strategy is proposed,The simulation experiments verify that the improved algorithm can effectively improve the localization and mapping accuracy of unmanned underwater vehicleThen,since each particle in the underwater SLAM algorithm based on random finite set carries an independent and complete grid map,The maps between the particles have great similarity,In order to make full use of the redundancy of the map between particles,a data structure ancestry tree is proposed.The data structure ancestry tree records the intergenerational succession between all particles,which reflects the motion relationship between particles and the inheritance relationship of map resources.By pruning and merging the tree,the size of the tree is maintained within a certain range.Finally,in the resampling process of the underwater SLAM algorithm based on random finite set,a distributed particle mapping method is proposed to deal with the large amount of map data flow and memory consumption.In this method,the two data structures of ancestry tree and distributed grid map are closely related.The data structure distributed grid map uses the relationship between particles,through an efficient map management method,to connect all particles with one grid map,and changes the mapping relationship between particles and grid map from one to one For many-to-one,the memory requirement for map maintenance is reduced,and the efficiency of map maintenance is improved.