Effective Map Matching Using Curve Tangents and Hidden Markov Model

A GPS sensor is utilized to determine the position of objects on earth and it is enormously beneficial to society. Despite the advances in GPS technologies, GPS measurement is still susceptible to errors, causing it to only highlight the coarse position of a place or an object. Map matching is the process of matching erroneous GPS sensor readings from a device to a road network. The prime objective of map matching is to rectify the positioning errors associated with GPS measurements. In this paper, we propose a novel map matching technique that utilizes Hidden Markov Model (HMM), tangent distance and some geometric properties of road segments. Specifically, to determine the transition probability of the HMM, we portray a road as a finite interchanging combination of straight lines, transition curves and circular curves, and craft an effective road geometry technique to differentiate circular circle from transition circle using the deflection angle. Besides, we utilize a set of points on each road segment and the tangent distance to compute the emission probability. Subsequently, we employ the Viterbi algorithm to find the most likely road segment to which a sequence of GPS readings correspond to. We conduct numerous experiments at different sampling rates using the Microsoft Seattle real dataset. Our experiments demonstrate that our technique yields a higher map matching accuracy than that of a state-of-the-art work.