Reproduce “MuNES: Multifloor Navigation Including Elevators and Stairs” on OpenResearch

About this paper

Donghwi Jung, Chan Kim, Jae-Kyung Cho, Seong-Woo Kim

Researchers from Seoul National University developed MuNES, a system that creates a single, accurate multi-floor map including explicit representations of stairs and elevators by using barometric pressure data for elevation estimation and floor-based loop detection. The system generates optimal multi-floor trajectories for mobile robots by applying the Multifloor Traveling Salesman Problem (MF-TSP) with a cost function that incorporates realistic factors like elevator waiting times.

Key insights

• Barometric pressure sensors offer a robust and reliable means for continuous altitude estimation and correcting z-axis drift, even in environments where LiDAR features are sparse or static, such as inside an elevator.
• Incorporating floor labels into loop detection significantly improves precision and reduces the search space, addressing the challenge of repetitive structures in multi-floor indoor environments.
• Optimal multi-floor trajectory planning necessitates a dynamic cost function that accounts for real-world operational delays, like elevator waiting times, allowing robots to make context-aware decisions between different inter-floor transition methods.

Summary by alphaXiv.