The Research On Robot Autonomous Navigation Algorithms Based On Incremental Smoothing And Mapping

Navigation, which is a critical issue for underwater vehicles, has been one of thehot spots that researchers are working on recently. Navigation and localization withhigh precision is vital for the safety, effective completion of missions of the vehiclesand effective use of the collected underwater data. The ability of autonomousnavigation can ensure that a mobile robot reaches the target location from a specifiedaccurate position, and completes the designated tasks in known or unknownenvironment. In order to realize the robot autonomous navigation ability,Simultaneous Localization and Mapping (for short SLAM) algorithm, as an importantmethod, is of the essence and indispensable. SLAM is the problem of estimating theposition of the robot and simultaneously building up a feature map of the environmentwhere the robot is moving based on the control input and observations without thehelp of prior environmental information map.Although SLAM plays a key role in robot navigation and exploration field, itsdeficiency in consistency, computational complexity and other aspects makes it notvery effective in the large range of environment. For the practical application of robotin most uncontrolled environments, the key problem is how to integrate the localsensor data into a consistent global map. This problem is very difficult to SLAM,because SLAM algorithm will estimate the position of the mobile robot and build up amap of environment only depending on the local measurements from sensors. Even avery tiny local error can lead to a great global error and cause the environment mapinconsistency with the accumulation of time. So incremental smoothing and mapping(for short iSAM) algorithm is introduced in this paper, which is a new algorithmaccess to the SLAM problems. It provides an efficient and exact solution by updatinga QR factorization of the naturally sparse smoothing information matrix based on afast incremental matrix factorization and can compute the full map and trajectory atany time. The paper firstly discusses the principles of the incremental smoothing andmapping algorithm and analyzes the realization process of the algorithm concretely.Through the Victoria Park dataset, it systematically assesses the iSAM algorithm andproves its validity. This paper gives a detailed description to the theory of dataassociation and studies all sort of common data association techniques. It alsopresents marginal covariances recovery for iSAM. Finally two simulationexperimental results verify the feasibility of iSAM algorithm.