An Algorithm And Implementation Of Simultaneous Localization And Mapping For Auv

Autonomous underwater vehicle (AUV), which is widely used in marine exploration and resource development, is a focus in the robotic research field today. And navigation is the core of the autonomous mobile robot-related technology. Navigation and localization information with high precision plays a decisive role in completing the mission safely and effectively. Simultaneous localization and mapping (SLAM) algorithm relies on self-contained sensors to extract the useful information to locate itself without the help of prior map information, and at the same time, construct the feature map of underwater environment. SLAM algorithm has great potential in underwater navigation, so it has been a focus in the field of autonomous mobile robot research for the past two decades.This paper has two main purposes. Firstly, it uses EKF-SLAM to implement AUV's localization and build environment maps according to the sensors data from the C-RANGER AUV. Secondly, it uses multithreading technology to meet the real-time requirement for the SLAM algorithm whose implementation is time-consuming.This paper mainly contains the following sections. The first section introduces the development and research condition of AUV and SLAM algorithm, and multithreading is proposed at the same time. Then, this paper explains the system model and the SLAM execution process on the C-RANGER platform. The result of simulation experiments done by MATLAB demonstrates that this method is convergent and the localization accuracy is higher than dead reckoning. Overall, it can meet the requirements of AUV navigation. The third section introduces the software architecture of the C-RANGER platform. The fourth section is the key point of this paper, which describes how to use multithreading to satisfy the real-time requirement of SLAM system. Firstly this section analyzes reasons why SLAM algorithm is so time-consuming, and the updating parts of sonar sensor and inertial sensors are put in one thread respectively so that they can run in parallel to meet the real-time requirement. The last section summarizes the paper and proposes the improvement as a reference for future work.