Abstract
Multi-hypothesis tracking (MHT) is one efficient way of global localization. One of the advantages of MHT is that pose hypotheses are compact and reliable since they are obtained not by random sampling, but by matching sensor data and map. Matching process usually takes adequate computation time for simple environments; however, when the environment is crowed by many objects and the sensor data is highly cluttered, we need tremendous time only for generating pose hypotheses. In this paper, Vector Pattern Matching method is presented for hypothesis generation, which shows great efficiency over traditional methods.
