Characterizing road segments using compass sensors to predict approaching bus stops

We devise an inexpensive and intuitive system for bus route navigation for locales where public transportation may serve as a prevalent mode of commute but where technologies that make arrival predictions through tracking vehicles in transit through GPS or other means do not exist. These systems typically require real-time monitoring of traffic variations. We provide a personalized approach where in a world of pervasive smart phone use, users may take advantage of sensor data to learn and personalize their bus routes, and alert them on time when a bus stop is approaching. We accomplish this through the development and implementation of two algorithms: 1) turn detection using on-board compass sensor of a smart phone, and 2) characterizing road segments in terms of turns and thereby predicting approaching bus stops. We conduct field experiments on a route with four selected bus stops in the town of Chapel Hill. Results show that the accuracy of turn detection and detection of approaching bus stops are 95.7% and 83%, respectively.