Robust Scale Estimation System for Monocular Mobile Robots Using Beacon-Based Structure from Motion

The precise estimation of scale in Structure from Motion (SfM) pipelines holds paramount significance for robotic systems, influencing their navigational capabilities, object manipulation, and decision-making processes. This paper presents an innovative prior-knowledge approach designed to address the challenge of scale ambiguity in monocular robots by strategically utilizing beacons positioned at known locations within the environment. Our methodology integrates well-established optimization techniques into a highly modular pipeline, offering adaptability to a spectrum of use cases and requirements. To validate the effectiveness of our approach, we conducted benchmarking experiments utilizing synthetic data (ICL-NUIM) and simulated data. The evaluation of our method on the ICL-NUIM dataset underscores its capability to correct the scale drift with comparable accuracy. The results highlight the potential of our approach to serve as a robust system across diverse scenarios, showcasing its viability for implementation in real-world applications.