Embodied AI Glossary中文Social Navigation
社交导航AdvancedA robot moving through crowds of people, reaching its goal while respecting human social norms.
Social navigation studies how a robot should move through environments with people in them, such as malls, stations, offices, and homes: not just avoiding collisions and reaching a destination, but moving in a way that feels natural and comfortable to the people around it, such as keeping an appropriate personal distance, not walking between two people talking, and yielding proactively. A 2023 survey by Mavrogiannis and colleagues in ACM Transactions on Human-Robot Interaction groups the core challenges into motion planning, behavior design, and evaluation, and notes that the field lacks a unified evaluation standard. Common methods include velocity-obstacle approaches such as ORCA, social-force models, and policies trained in simulation with reinforcement learning or imitation learning. Meta's Habitat 3.0 also defines “find and follow a person” as a social navigation task.
ExampleIn Habitat 3.0's social navigation task, a robot must find a humanoid avatar in a previously unseen home scene and follow it, maintaining a safety distance of 1 to 2 meters throughout; any collision counts as a failure.
- Also called
- Social Robot Navigation
- Related
- Navigation · Human-Robot Interaction · Velocity Obstacles · Habitat · Embodied Safety
- Sources
- Core Challenges of Social Robot Navigation: A Survey (ACM THRI, 2023)
Habitat 3.0: A Co-Habitat for Humans, Avatars and Robots