A Hybrid Deep Sea Navigation System of LBL/DR Integration Based on UKF and PSO-SVM

来源：机器人杂志2015年第5期北京时间：

作者：LIU Ben, LIU Kaizhou, WANG Yanyan, ZHAO Yang, CUI Shengguo, WANG Xiaohui

摘要：In order to improve the navigation accuracy of human occupied vehicle (HOV) precisely and efficiently, aninnovative hybrid approach based on unscented Kalman filter (UKF) and support vector machine (SVM) is proposed tofuse integrated navigation data. HOV is generally equipped with long baseline (LBL) acoustic positioning system and deadreckoning (DR) as an integrated navigation system. UKF is adopted to estimate the state of the dynamic model becauseof its good performance in filtering nonlinear problems. An accurate and stable filtering result can be obtained when bothLBL and DR are online. At the same time, SVM is utilized to train DR information with the result when LBL outrages, andthe particle swarm optimization (PSO) algorithm is employed for SVM parameters optimization. Therefore, the integratednavigation system can maintain a good performance when the LBL is off-line. Simulation results with the real navigationdata of Jiaolong HOV show that the methodology proposed here is able to meet the needs of HOV application.

关键词：unscented Kalman filter (UKF), particle swarm optimization (PSO), support vector machine (SVM), deep seanavigation system, human occupied vehicle (HOV)