The Comparison of Gauss-Newton and Dog-Leg in iSAM for AUV

Abstract

Autonomous underwater vehicle (AUV) plays an important role in complex underwater environments such as marine surveys, due to their flexibility and autonomy. One of the main challenge for AUVs is autonomous navigation technology for the safety and effectiveness of AUV missions. Simultaneous localization and mapping (SLAM) plays an important role in autonomous navigation. However, a suitable optimization algorithm can not only improve the performance of navigation, but also can save running time for AUV. In this paper, we adopted two algorithms of optimization in Incremental Smoothing and mapping (iSAM) for AUV. Several sea trials were performed at Jiaozhou Bay, Qingdao City, Shandong Province. Compared with the result of trajectory using Dog-Leg for optimization, the Gauss-Newton track is more close to the ground truth. At the same time, RMSE of Gauss-Newton trajectory and the computational time is superior obviously to that of Dog-Leg. We consider the Gauss-Newton method is more suitable to perform optimization in iSAM for AUV.