3D Traversability Map Generation for Mobile Robots Based on Point Cloud

Abstract

Analysing the traversability from the point cloud input is a critical task for mobile robot navigation. However, there is no mature method for mobile robots to exploring complicated 3D environments, especially those with multi-layered roads. This paper proposes a new traversability map based on the point cloud input, and the map will help analyze where the mobile robot can move in the whole space. The proposed 3D traversability map consists of nodes and edges generated according to the mobile robot’s geometry and driving performance. Nodes represent where the robot can be placed, and edges indicate the cost of moving to another node. The traversability map is stored in digraph form and is convenient to be applied to path planning. Experiments show the successful results of traversability map generation and path planning with D algorithm in different situations.