| With the development of ocean,underwater robots can work instead of human beings in a complex and harsh environment.How to carry out autonomous work is an urgent problem to be dealt with,and positioning is the key to realize autonomous work of robots.Based on the research of underwater positioning methods at home and abroad,the multi-sensor fusion positioning algorithm for underwater robot is studied deeply in this paper.Compared to the more common land-based robots and drones,underwater robots face unique challenges.The underwater environment is more complex than the onshore environment,and there are many challenges such as the inability to use GPS and limited visibility.These challenges make it difficult for the traditional positioning method using a single sensor to achieve the desired effect underwater.To solve these problems,this paper proposes a positioning scheme that integrates the improved visual odometer with sonar and inertial sensor.In this scheme,Firstly,through the exploration and comparison of the principle,structure and front-end methods of Simultaneous Localization and Mapping algorithm,the front-end method suitable for underwater environment is selected to provide theoretical and practical basis for the front-end architecture of the algorithm.Then,in view of the limited visibility of underwater environment,an image preprocessing module was added into the visual odometer,and the result of feature extraction in the front end was significantly improved by the improved image enhancement algorithm and denoising operation in series.Aiming at the back end,a fusion method of vision,sonar and inertial sensor based on nonlinear optimization is proposed.By constructing error terms,the vision,sonar and inertial information are integrated organically,which improves the localization performance and robustness of the vision-based SLAM algorithm in underwater.Finally,a simulation experiment is designed to verify the effectiveness and feasibility of the proposed method in underwater environment. |
