The Study On SLAM Based On Particle Filter And Consistency Analysis For AUV

Autonomous Underwater Vehicles (AUV), also known as unmanned underwatervehicles, can be used to perform underwater survey missions such as detecting andmapping submerged wrecks, rocks, and obstructions that pose a hazard to navigationfor commercial and recreational vessels. The AUV conducts its survey missionwithout operator intervention. When a mission is complete, the AUV will return to apre-programmed location and the data collected can be downloaded and processed inthe same way as data collected by shipboard systems. AUVs can be equipped with awide variety of oceanographic sensors or sonar systems. AUVs development is theinevitable tendency of underwater vehicle in the future, many countries begin to payhigh attention to the research of AUVs and have put into a lot of money and time.With the widely application of AUVs, autonomous navigation technology has gotmore and more attention. The capacity of robot localization and map building at thesame time, namely the simultaneous localization and mapping (SLAM)，is the key ofrealizing truly autonomous navigation. In latest years, people have been researchingand improving the navigation algorithm all the time. Among all of those solutions, theimproving of the efficiency and coherence of the algorithm has extremely importanteffects on the accuracy of the whole algorithm. Over the past years, particle filtershave been applied with great success to a variety of state estimation problems，but itstill remains problems such as inconsistency.In this paper, we firstly introduced several popular SLAM algorithm and theirdevelopment history. Secondly, our own AUV platform C-Ranger and the sensors onit are presented. Then, we elaborate the basic Particle Filter algorithm and ourimprovement based on the basic algorithm. At last, the effectiveness of the improvedalgorithm is verified though the simulation experiment, the Victory Park dataset andthe Sea Trial, in which the Sea Trial is carried out utilizing our own AUV platform C-Ranger and is performed at Tuandao Bay in Qingdao.In this paper, we propose an improved particle filter(PF) using theFirst-Estimates Jacobian(FEJ) in the landmarks estimation of the particle filter toimprove the accuracy of landmarks estimation, which contributes to build a consistentmap and enhance the consistency of the entire particle filter. Meanwhile, we useeffective number of particles in the resampling procedure to reduce the problem ofparticle degeneracy. The consistency of the improved algorithm is verified through thenormalised estimation error squared(NEES) and we compare it with the traditionalalgorithm to validate the efficiency of the new one.