Ziegler P, Franke J, Reitelshöfer S (2025)
Publication Type: Conference contribution, Conference Contribution
Publication year: 2025
Publisher: Institute of Electrical and Electronics Engineers Inc.
Conference Proceedings Title: 2025 European Conference on Mobile Robots, ECMR 2025 - Proceedings
Event location: Padua, ITA
ISBN: 9798331527051
DOI: 10.1109/ECMR65884.2025.11163160
In structured indoor environments, autonomous mobile systems (AMS) are becoming increasingly predominant. Their ability to dynamically adapt navigation paths according to changing environmental conditions offers significant advantages, for example, in intralogistics operations. State-of-the-art systems rely on laser-based environment perception to navigate their surroundings. This approach currently excels due to reliable deterministic algorithms that ensure sufficient safety levels. However, as AMS expand into unstructured environments, such as fields or construction sites, more adaptable perception solutions are required. To address this challenge, we propose a novel neural network architecture that integrates stereovision data for robust semantic segmentation of environment obstacles, enhancing perception in diverse environments. This architecture enables accurate and real-time obstacle detection even in complex, changing scenarios. Additionally, to mitigate the challenges posed by limited training data availability, we utilize an automated data annotation pipeline that facilitates rapid adaptation to new environments with reduced human oversight, improving the scalability and economic viability of our data-driven perception model. The evaluation demonstrates that our mid-level fusion approach retains accuracy, especially in challenging conditions, with the multimodal model demonstrating an Intersection over Union of 0.905 outperforming the standalone architectures at 0.834 and 0.745 respectively.
APA:
Ziegler, P., Franke, J., & Reitelshöfer, S. (2025). Mobile Robot Environment Perception: Adaptable Real-Time Neural Network Architecture for RGB-D Ground Segmentation*. In Antonios Gasteratos, Nicola Bellotto, Stefano Tortora (Eds.), 2025 European Conference on Mobile Robots, ECMR 2025 - Proceedings. Padua, ITA: Institute of Electrical and Electronics Engineers Inc..
MLA:
Ziegler, Patrick, Jörg Franke, and Sebastian Reitelshöfer. "Mobile Robot Environment Perception: Adaptable Real-Time Neural Network Architecture for RGB-D Ground Segmentation*." Proceedings of the 12th European Conference on Mobile Robots, ECMR 2025, Padua, ITA Ed. Antonios Gasteratos, Nicola Bellotto, Stefano Tortora, Institute of Electrical and Electronics Engineers Inc., 2025.
BibTeX: Download