Vehicle specific robust traversability indices using roadmaps on 3D pointclouds

摘要

Roads undergo rapid deterioration due to various economic, social and natural reasons. The uncertainty and unavailability of road safety information has become a major issue in reliably transporting goods, movement of heavy machinery, transport of people and materials to/from disaster areas and in even giving plain guarantees for safe traversability of a road that is good on paper. In this paper, we propose a novel approach towards reducing this uncertainty. Instead of asking qualitative questions about whether a road network is generally traversable for a general class of vehicles, we propose a solution that gives precise quantitative answers as a distance modulated roadmap and confirms whether a particular segment of the road is traversable for a particular type of vehicle. In this paper, we extend on our previous work Khan et al. (2016) of traversability analysis by empirically proving that our framework is robust to shape and size of the roadmap graph. We show that obstacle size, count, and location can have varying effects on traversability of different vehicles. We also propose a new Road Safety Index (RSI) which is an extension of our previously proposed index, Road Traversability Index (RTI). RSI takes into account multiple disjoint narrow passages that belong to the same connected component of the graph. Lastly, we present application scenarios where our roadmap based traversability can be used to assess the safety of the road for a particular vehicle.