Vehicle Destination Prediction Using GPS Tracking: Analysis Using Velocity Based Particle Filters

Description

Road navigation has gained increased importance in recent times with an increased number of vehicles making use of GPS navigators. As a result, focus has been placed on services that can be offered to drivers by prediction of routes they might take to their final destination. The analysis here predicts the destination of a driver given an initial starting point and regular position coordinates supplied by a GPS sensor. The system is simulated with the help of an a priori knowledge of the road network and its cost distributions supplemented by the individual driver's history and road preference. The novelty of the system lies in the manner in which the driver's destination probabilities are calculated and varied. By the use of velocity-based particle filters combined with a novel idea for the set up and analysis of road link cost distributions, an algorithm is created which is seen to perform to a high standard. The destination of the driver is identified as one of a few highly probable candidates within the first 50 percent of journey traveled. By 75-80% of the journey, the destination node is pinpointed with certainty.