Brain-inspired spatial intelligence for embodied agents
Shouwei Ruan, Liyuan Wang, Caixin Kang, Qihui Zhu, Songming Liu, Xingxing Wei, Hang Su · 2026
Read Brain-inspired spatial intelligence for embodied agents by Shouwei Ruan, Liyuan Wang, Caixin Kang, Qihui Zhu, Songming Liu, Xingxing Wei, Hang Su (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13454140.txt.
Description
Shouwei Ruan and six co-authors give embodied AI agents a brain inspired way to hold space in memory, rather than reacting frame by frame. Brains store space in three layers: landmarks, routes, and map like survey knowledge. Their 2026 framework, BSC-Nav, builds map style memory from an agent's own path and cues, then pulls the right piece back when a goal is named.
Paired with large multimodal models, the system topped prior scores on several navigation tests. Success weighted by path length jumped from 17.6 to 44.9 percent on instance level navigation, and from 42.7 to 53.1 percent on long instruction following with no task training. The same memory stack drove varied behaviors on a real physical robot. The authors present structured spatial memory as a scalable path to general robot intelligence.
Catalog record
| CREATOR | Shouwei Ruan, Liyuan Wang, Caixin Kang, Qihui Zhu, Songming Liu, Xingxing Wei, Hang Su |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 87121 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |