Terrain Based GPS Independent Lane-Level Vehicle Localization Using Particle Filter and Dead Reckoning

The need of accurate and reliable positioning in various location-aware safety critical transportation applications is increasing day by day. The Global Positioning System (GPS) is not able to provide lane-level vehicle localization without the aid of differential corrections. It also suffers from signal outages in urban areas resulting in a complete loss of location information. Therefore, GPS independent localization methods are now being developed. In this domain, inertial sensors along with a terrain map have been successfully deployed to achieve sub-meter level accuracy in the longitudinal direction of the vehicle in an urban environment. However, lateral localization of the vehicle with good accuracy and computational efficiency remains a challenging topic. Existing algorithms are computationally intensive, and do not provide location information during the process of lane change by the vehicle. This information is very crucial as the risk of potential conflict with nearby vehicles is higher during lane changes. In this paper, we present a computationally efficient method for achieving lane-level localization in a multi-lane scenario by combining the particle filter with dead- reckoning. The particle filter provides the location information about a single lane while location information during the lane change maneuvers is provided by dead-reckoning. Lane- change maneuvers are detected by constantly observing the yaw rate of the vehicle. Developing a computationally efficient algorithm enables the GPS independent localization algorithm to be run on low cost micro-controllers making its deployment feasible for packaged devices. Experiments performed on an instrumented vehicle show the superiority of the proposed algorithm on the existing ones.