VAP-TAMP wins IROS best paper: VLMs help robots handle surprises
chris_j_paxton · x · 2026-10-14
VAP-TAMP, a planning and situation-handling framework by Oloo, Zhang, Paxton et al., won the best paper award at the IROS full-shift robot co-workers workshop.
- Problem: Robots frequently hit unforeseen situations during execution — a half-open door, a lemon half falling off the plate — limiting long-horizon autonomy.
- Approach: Combines task and motion planning with vision-language models, using action knowledge to strategically prompt VLMs for active view selection and situation assessment while reasoning over scene graphs.
- Results: Validated in simulation and on a real mobile manipulation platform, robustly handling failures and out-of-distribution scenarios.
Project page: vap-tamp.github.io/vap-tamp
More from Embodied
- Nvidia's TacitVLA matches CoT performance at up to 3.4x lower latency — arankomatsuzaki · 2026-10-14
- SlashVibe, a physical controller for AI agents, hits Kickstarter October 27 — sujingshen · 2026-10-14
- Fchollet: Autoregressive Sampling Alone Can't Solve Real-World Robotics — fchollet · 2026-10-14
- AlphaNIST: binary-reward RL trains one policy to 98.2% success on all 20 NIST assembly tasks — chris_j_paxton · 2026-10-14
- CUHK's Orchestra robot framework completes 7.35-step subtask sequences — CUOHK · 2026-10-14
- Robot Learns Hand Skills From 1M Hours of First-Person Video, Generalizes to Unseen Tasks — Scobleizer · 2026-10-14