New benchmark tests whether home robots navigate more cautiously around babies than trash bins
SeoulNatlUniv · hf · 2026-10-28
Seoul National University researchers released a benchmark and metric evaluating whether robot navigation policies adapt motion to obstacle safety semantics. Key idea: not colliding isn't enough—a safe robot keeps more clearance and moves slower near a baby than a trash bin, slowing further as it approaches.
The benchmark simulates policies (VoxPoser, PILOT, VLA) across models on identical routes with swapped obstacles. Results reveal a real gap: some models like π0.5-OpenTau collide with the child, while the best models achieve 0.45m min clearance near a child vs 0.24m near a cardboard box—showing semantic caution often fails to translate into safe motion.
More from Embodied
- PDRVLA adds physical dynamics tokens to VLA models, topping LIBERO with 99.3% — Hanqing Wang · 2026-10-28
- Alibaba DAMO releases RynnWorld-Latent: cross-embodiment world model pretrained on 553.5M frames — Alibaba-DAMO-Academy · 2026-10-28
- Developer hooks GPT into a Hasselblad: focus, shutter and chat, open-sourced — rrRadium · 2026-10-28
- Robots Coming for Your Jobs: A Playful Reaction to a Demo — alysha_lobo · 2026-10-28
- Early adopter revisits AI hardware: still meh, runs hot without opening apps — alexmacgregor__ · 2026-10-28
- RobotAgentBench debuts to test open-world robot autonomy: manipulation ≠ autonomy — HennyJieCC · 2026-10-28