Map-based lane identification and prediction for autonomous vehicles

Abstract

A novel map-based lane identification and prediction algorithm is developed to characterize areas around an autonomous robot as it travels in an obstacle strewn and rugged roadway environment. The implementation of this algorithm employs probabilistic and heuristic methods to improve the placement of lane features, whose location is uncertain due do to vehicle motion and sensor data ambiguity. The resulting map can be effectively used for local and regional path planning and navigation. The algorithm uses data acquired from a LIDAR, compass, GPS, wheel encoders, and camera images.