ACME Dataset: Cross-Cultural Social Navigation for Robots
A team of researchers has introduced ACME, an extensive multi-modal dataset designed for the navigation of social robots. This dataset was gathered from 8 locations across 5 different countries, utilizing 7 distinct robot embodiments. It comprises 29.35 hours of data recorded onboard the robots and 43.5 hours of tracking pedestrian movements from above, emphasizing goal-oriented behavior in intricate social situations, particularly highlighting robot interactions with crowds through verbal communication. The initiative seeks to fill the gap in diversity present in current datasets concerning culture, geography, and the dynamics of human-robot interaction.
Key facts
- ACME stands for A Cross-cultural, Multi-Embodiment dataset for social navigation.
- Data collected across 8 sites in 5 countries.
- Used 7 robot embodiments.
- 29.35 hours of onboard robot data.
- 43.5 hours of overhead pedestrian tracking data.
- Focuses on goal-driven social navigation with robot speech.
- Published on arXiv with ID 2607.21964v1.
- Designed to improve robot navigation policies and human behavior prediction.
Entities
Institutions
- arXiv