A new approach to map-assisted Bayesian tracking filtering

摘要: Abstract This paper presents a new non-linear filter designed to track targets following road network, taking advantage of the map information. The algorithm is based on Bayesian Multiple Hypotheses modelling movement process, postulating and evaluating different hypotheses segments being followed by target after junctions. Then, along-road tracking carried out, for each hypothesis, longitudinal IMM capable movements along straight roads, circular segments, generic curvilinear defined through Bezier curves. also includes lateral drift estimator, which tracks motion with respect axis, be able estimate piloting error especially in wide roads. completely describes associated measurement preprocessing procedures, comparative evaluation proposed other filtering methods literature.