A LiDAR aiding ambiguity resolution method using fuzzy one-to-many feature matching

Despite the high-precision performance of GNSS real-time kinematic (RTK) in many cases, large noises in pseudo-range measurements or harsh signal environments still impact float ambiguity estimation in kinematic localization, which leads to ambiguity-fixed failure and worse positioning results. To improve RTK ambiguity resolution (AR) performance further, multi-sensor fusion technique is a feasible option. Light detection and ranging (LiDAR)-based localization is a good complementary method to GNSS. Tight integration of GNSS RTK and LiDAR adds new information to satellite measurements, thus improving float ambiguity estimation and then improving integer AR. In this work, a LiDAR aiding single-frequency single-epoch GPS + BDS RTK was proposed and investigated by theoretical analysis and performance assessment. Considering LiDAR-based localization failure because of ambiguous and repetitive landmarks, a fuzzy one-to-many feature-matching method was proposed to find a series of sequences including all possible relative positions to landmarks. Then, the standard RTK method was tightly combined with the possible positions from each sequence to find the most accurate position estimation. Experimental results proved the superiority of our method over the standard RTK method in all aspects of success rate, fixed rate and positioning accuracy. In specific, our method achieved centimeter-level position accuracy with 100% fixed rate in the urban environment, while the standard GPS + BDS RTK obtained decimeter-level accuracy with 26.84% fixed rate. In the high occlusion environment, our method had centimeter-level accuracy with a fixed rate of 96.31%, comparing a meter-level accuracy and a fixed rate of 7.65% of standard GPS + BDS RTK method.