Safety through perception: Multi-modal traversability analysis in rough outdoor environments

M. Breitfuß, M. Schöberl, J. Fottner

Research output: Contribution to journalConference articlepeer-review

1 Scopus citations


One of the most crucial functionalities of mobile indoor and outdoor robots is the ability to understand the local environment in a reasonable manner. Based on an intelligent perception of the systems' surroundings, a secure input for further autonomous functionalities like path planning or navigation can be guaranteed. Especially in the domain of off-road vehicles and machinery difficulties concerning a meaningful comprehension of the environment increase, as within this field of application the surroundings are rough and unstructured, implicating unpredictable situations for perception algorithms. Therefore, procedures to analyse the traversability of off-road terrain have to be both, robust in order to ensure a reasonable understanding of the environment under different circumstances and flexible to be adaptable to unknown application areas. In this paper we propose a method satisfying these requirements. Our approach is based on an efficient way of combining geometric and visual features of the surroundings, leading to a real-time capable traversability analysis for autonomous off-road vehicles. The evaluation of the developed system has been conducted in virtual tests within a simulation as well as in real-world situations on a prototypical vehicle.

Original languageEnglish
Pages (from-to)223-228
Number of pages6
JournalIFAC Proceedings Volumes (IFAC-PapersOnline)
Issue number1
StatePublished - 2021
Event17th IFAC Symposium on Information Control Problems in Manufacturing INCOM 2021 - Budapest, Hungary
Duration: 7 Jun 20219 Jun 2021


  • Autonomous systems
  • Decision making
  • Safe environmental perception
  • Semantic segmentation
  • Traversability analysis


Dive into the research topics of 'Safety through perception: Multi-modal traversability analysis in rough outdoor environments'. Together they form a unique fingerprint.

Cite this