Publication
traversability_generator3d: A C++ library for 3D traversability estimation from MLS maps
Haider Khan Lodhi; Christoph Hertzberg; Arne Böckmann; Janosch Machowinski
In: Journal of Open Source Software (JOSS), Vol. 11, No. 118, Pages nn-nn, The Open Journal, 2026.
Abstract
traversability_generator3d is a C++ library for 3D traversability estimation based on Multi-Level Surface (MLS) maps. Unlike conventional 2.5D elevation maps that store only a single surface per grid cell, MLS representations preserve multiple surface hypotheses, enabling robust reasoning in complex environments with overlapping terrain layers such as rubble, vegetation, bridges, or multi-level structures. The library converts MLS maps into robot-specific 3D traversability maps by performing plane fitting, slope and step-height analysis, orientation-dependent motion constraints, and footprint-aware obstacle inflation. In addition to geometric reasoning, it supports soil-aware traversability by integrating probabilistic soil semantics with configurable cost models and forbidden terrain types. The resulting TraversabilityMap3d provides rich node classifications and cost information suitable for autonomous navigation and exploration. traversability_generator3d has been applied in real-world robotic systems for construction, disaster response, search-and-rescue, and planetary exploration scenarios.