Abstract:

The result of data association is affected by the state of features in the environment. In order to analyze the influence, two kinds of simulation scenes are set up, where feature pairs are arranged at different intervals. Four kinds of data association methods, individual compatibility nearest neighbor (ICNN), sequential compatibility nearest neighbor（SCNN, joint compatibility（JC） and joint maximum likelihood （JML） are tested in different situations for the comparison of feature association. The results show that the featurelocations and the interval between features have important effects on the association method. The worse results can be gotten for the most association method when the interval is not large, but good results can be gotten when the interval is large, and each method has its advantages and disadvantages in different situations. The results are of significance in applications to the analysis and selection of the method of data association in simultaneous localization and mapping (SLAM).