Robot navigation handles crowds and sidewalks at the same time
SNAP: Semantic Navigation Among Pedestrians
Robotics
Summary
Moving safely through busy city streets is hard for robots because they need to avoid bumping into people while also staying on sidewalks and following social rules. The authors created a new way for robots to learn navigation that looks at both where pedestrians are and where sidewalks and roads are, all at once. Their method uses a special map from above that shows these details, and it helps robots move naturally among people without needing to track every person’s exact position. Their approach works better than older methods and can be used directly on robots that walk on legs in real busy environments.
What this means in practice
- •For robotics engineers: Develop walking robots that can navigate crowded urban sidewalks without colliding with pedestrians or stepping off designated spaces.
- •For autonomous delivery teams: Enhance delivery robots to move safely and socially appropriately through pedestrian-heavy city areas using a single overhead semantic map.$Commercial implications: Enables commercial robots to reliably navigate busy sidewalks, addressing safety and social norms essential for product adoption.
Authors
Jakob Wolfram, Pascal Roth, Fan Yang, Cesar Cadena, Marco Hutter
Abstract
Navigating crowded urban environments requires robots to avoid pedestrians while adhering to social norms such as remaining on sidewalks. Existing approaches typically address either crowd avoidance or semantic path planning in isolation. We propose a reinforcement learning (RL) navigation policy that jointly reasons about semantic terrain and dynamic obstacles from a single semantically annotated bird's-eye-view (BEV) map, without relying on explicit human-state estimation. Our policy outperforms established crowd-navigation and semantic-aware baselines and transfers zero-shot to a legged robot, demonstrating socially compliant navigation in highly dynamic real-world scenarios.