Similar but Different: A Survey of Ground Segmentation and Traversability Estimation for Terrestrial Robots

TL;DR

This work clarifies the long-running confusion between ground segmentation and traversability estimation for terrestrial robots by defining ground segmentation as a perception-level task that partitions input data into ground and non-ground regions, and traversability estimation as a cognition-level task that assesses where a robot can safely move. It surveys ground segmentation methods—elevation-based, geometric model fitting, and regression-based—alongside traversability approaches that include terrain classification and traversability analysis, highlighting how platform dynamics shape outcomes. A four-criteria framework (maneuverability, position, negative obstacles, deformable objects) is proposed to distinguish the two concepts and explain their differing dependencies. The findings stress that ground is largely platform- and position-invariant, while traversability is sensitive to robot capabilities and context, guiding researchers to align perception and planning components with the correct semantic distinctions for robust navigation in diverse environments.

Abstract

With the increasing demand for mobile robots and autonomous vehicles, several approaches for long-term robot navigation have been proposed. Among these techniques, ground segmentation and traversability estimation play important roles in perception and path planning, respectively. Even though these two techniques appear similar, their objectives are different. Ground segmentation divides data into ground and non-ground elements; thus, it is used as a preprocessing stage to extract objects of interest by rejecting ground points. In contrast, traversability estimation identifies and comprehends areas in which robots can move safely. Nevertheless, some researchers use these terms without clear distinction, leading to misunderstanding the two concepts. Therefore, in this study, we survey related literature and clearly distinguish ground and traversable regions considering four aspects: a) maneuverability of robot platforms, b) position of a robot in the surroundings, c) subset relation of negative obstacles, and d) subset relation of deformable objects.

Figures (8)

• Figure 1: Diagram of robot navigation with ground segmentation and traversability estimation. Note that ground segmentation is a technique for perceiving and mapping surroundings, whereas traversability estimation is aimed at cognition and motion control (best viewed in color).
• Figure 2: Subset relation between ground and traversable regions depending on the complexity of environments considering wheeled mobile robots. Note that in rough terrain environments, some parts of traversable regions may not be included in the ground owing to deformable objects (best viewed in color).
• Figure 3: Example of estimated ground and non-ground points by the ground segmentation approach, Patchwork++ lee2022patchworkpp. The green and red points denote the estimated ground and non-ground points, respectively (best viewed in color).
• Figure 4: (a)-(b) Examples of ground segmentation as a preprocessing step from egocentric and mapcentric perspectives, respectively. (a) A result of above-ground segmentation: ground segmentation is applied to reject ground points (gray points) before the main algorithm and then followed by above-ground segmentation oh2022travel. Points with the same color indicate that these points are segmented into the same object. (b) Before and after the application of map-level ground segmentation. By leaving non-ground points, ground segmentation allows a robot to perform better localization (best viewed in color).
• Figure 5: (a) Visual description of the difficulty of traversability, ranging from green to red. The green color represents the regions that are easy to traverse, whereas the red color represents the regions that are impossible to traverse. (b) Example of path planning lee2021real. The path candidates (fan-shaped yellow lines around the frame) are not generated near the trunks of trees (orange circle) by considering the traversability (best viewed in color).