Metric mapping and topo-metric graph learning of urban road network

Abstract

A road map serves as a model of the road network, which is especially desired for a vehicle performing autonomous navigation in urban road environment. This paper first introduces a metric mapping algorithm for urban roads, which generates an occupancy grid map of road surfaces and boundaries. Based on the metric map, we further propose an approach to extract a topo-metric graph which captures both topological and metric information of the road network. As a detailed model of the urban roads, the metric map can be used for obstacle avoidance and local path planning, while the topo-metric graph as a compact representation that can be used for some high-level reasoning processes. Our proposed algorithms are tested in real experiments, and have shown good results.