Dead reckoning using road networks

Abstract

Accurate vehicle position tracking in environments with limited Global Positioning System (GPS) availability remains a significant challenge in navigation systems. This thesis presents a dead reckoning algorithm that fuses data from accelerometers, gyroscopes, and speed sensors using a Kalman filter, supplemented by map-based corrections. By integrating known road network constraints and implementing a scoring system based on dynamic time warping, the algorithm enhances positioning accuracy during GPS outages. Extensive benchmarking across various road types and outage durations demonstrates strong performance on highways and country roads, with up to 77% accurate route prediction in short outages. The system also includes a backtracking mechanism to correct mispredictions, although limitations in scoring and prediction drift remain. Results suggest that context-aware parameter tuning and improved scoring methods could further enhance robustness in urban environments and extended outages.