Markerless Robot Detection and 6D Pose Estimation for Multi-Agent SLAM
Markus Rueggeberg, Maximilian Ulmer, Maximilian Durner, Wout Boerdijk, Marcus Gerhard Müller, Rudolph Triebel, Riccardo Giubilato
AI summary
Problem
Reliable inter-robot data association in multi-agent SLAM is hindered by perceptual aliasing, extreme lighting, and perspective differences, causing traditional fiducial markers to fail at long ranges or in unstructured terrains.
Approach
The authors embed a deep learning pipeline that detects robots and estimates their 6D pose directly from camera images using prior CAD shape knowledge, trained on diverse synthetic data and integrated into a decentralized SLAM system.
Key results
- First markerless mutual detection method using prior robot geometry without fiducial markers
- Successful integration of markerless pose estimation into a decentralized multi-robot SLAM framework
- Up to 402% increase in maximum detection distance and 87% more detections in field campaigns
- Up to 32% reduction in trajectory error and 55% shorter open-loop durations compared to AprilTag baselines
Why it matters
Provides a robust, hardware-free solution for collaborative mapping and localization of planetary rovers and outdoor multi-robot teams operating in challenging, unstructured environments.
Abstract
The capability of multi-robot SLAM approaches to merge localization history and maps from different observers is often challenged by the difficulty in establishing data asso- ciation. Loop closure detection between perceptual inputs of different robotic agents is easily compromised in the context of perceptual aliasing, or when perspectives differ significantly. For this reason, direct mutual observation among robots is a powerful way to connect partial SLAM graphs, but often relies on the presence of calibrated arrays of fiducial markers (e.g., AprilTag arrays), which severely limits the range of observa- tions and frequently fails under sharp lighting conditions, e.g., reflections or overexposure. In this work, we propose a novel solution to this problem leveraging recent advances in Deep- Learning-based 6D pose estimation. We feature markerless pose estimation as part of a decentralized multi-robot SLAM system and demonstrate the benefit to the relative localization accuracy among the robotic team. The solution is validated experimentally on data recorded in a test field campaign on a planetary analogous environment.