Physical AI Radar · 2026-09-28
A daily radar on the Physical AI frontier: eight lanes, trilingual digests, traceable evidence.
- Generated:
2026-09-27 23:40 UTC - Window:
papers 2026-09-24 → 2026-09-28 · posts 2026-08-29 → 2026-09-28 · no repeats within 30 days (UTC) - Evidence tags: O = first-party official · R = paper/preprint · M = media/secondary
Today’s picks
1. Beyond Binary Success: Sample-Efficient and Statistically Rigorous Robot Policy Comparison
- Lane: Simulation & evaluation (sim2real / closed-loop) ｜ Evidence:
[R]｜ Source: Toyota Research Institute ｜2026-09-09 - Key numbers:
70%(“we demonstrate up to 70% reduction in evaluation”) ·50%(“methods and up to 50% reduction compared to”) - Signals:
quantitative results·real-robot experiments - Why it matters
AI draft: Teams that compare policies on a handful of hardware rollouts may be able to cut that budget: the preprint self-reports up to 70% less evaluation and flags real-robot experiments, but check which baselines that figure is measured against first.
Generalist robot manipulation policies are becoming increasingly capable, but are limited in evaluation to a small number of hardware rollouts.
2. Using Non-Expert Data to Robustify Imitation Learning via Offline Reinforcement Learning
- Lane: Data engine (collection / curation / scaling) ｜ Evidence:
[R]｜ Source: Toyota Research Institute ｜2026-09-09 - Signals:
real-robot experiments - Why it matters
AI draft: If your imitation-learning pipeline keeps only expert demonstrations, this preprint argues the rest can be put to work through offline RL; it flags real-robot experiments but shows no figures here, so treat it as a lead on whether to stop discarding that data.
Imitation learning has proven effective for training robots to perform complex tasks from expert human demonstrations.
3. TACTIC: Tactile and Vision Conditioned Contact-Centric Control for Whole-Arm Manipulation
- Lane: Simulation & evaluation (sim2real / closed-loop) ｜ Evidence:
[R]｜ Source: Toyota Research Institute ｜2026-09-09 - Signals:
real-robot experiments - Why it matters
AI draft: When a task loads the whole arm rather than the gripper, with contacts forming, sliding and breaking across several links, this preprint offers a control approach with real-robot experiments; no figures appear here, so do not fix a sensing design on it yet.
Whole-arm manipulation involves direct contact with the environment while the robot completes a task by distributing contact across multiple links as contacts form, slide, and break.
4. ダイフクが物流倉庫の完全自動化へヒューマノイドを開発、2030年に実証実験
- Lane: Embodiment & supply chain (hardware / cost) ｜ Evidence:
[M]｜ Source: MONOist (ITmedia) ｜2026-09-27 - Why it matters
AI draft: For warehouse buyers, what Daifuku actually showed was automated storage combined with AMRs; the humanoid is a stated development direction in a secondary report with no figures, so it should not enter a procurement shortlist yet.
ダイフクは、「国際物流総合展2026」において、自動倉庫やAMRを組み合わせた物流倉庫向けの自動化ソリューションを披露。併せて、物流倉庫の完全自動化の鍵となるヒューマノイド開発の方向性についても言及した。
Frontier baseline (curated)
1. ISO 25785-1 (under development): safety for dynamically stable industrial mobile robots
- Lane: Safety, permissions & compliance ｜ Evidence:
[O]｜ Source: ISO ｜2026 - Key numbers:
committee stage as of 2026-09·covers legged / self-balancing robots - Why it matters: Humanoids break the fixed-base premise of ISO 10218; until the dedicated standard lands, safety rests on architecture and runtime guarantees.
2. EgoScale: scaling dexterous manipulation with diverse egocentric human data
- Lane: Data engine (collection / curation / scaling) ｜ Evidence:
[R]｜ Source: arXiv 2602.16710 ｜2026 - Key numbers:
44k hours of egocentric video·+54% average success vs no-pretraining on a 22-DoF hand·transfers to lower-DoF hands - Why it matters: Large-scale human motion is shown to be a reusable, embodiment-agnostic motor prior, so dexterous hands do not start from zero.
3. Z-1: efficient RL post-training for flow-based VLA models
- Lane: Training & self-improvement (RL / post-training) ｜ Evidence:
[R]｜ Source: arXiv 2606.31846 ｜2026 - Key numbers:
80.6% average success on 24 RoboCasa tasks·+13.2 points over SFT init·public demos only - Why it matters: Systematic GRPO post-training delivers double-digit gains without extra private demonstrations, breaking the default belief that only more data helps.
4. PL-CBF: finite-horizon safety at runtime via parallel fallback rollouts
- Lane: Safety, permissions & compliance ｜ Evidence:
[R]｜ Source: arXiv 2605.16588 ｜2026 - Key numbers:
library of fallback policies·least-invasive safe mode·QP minimally modifies the nominal policy - Why it matters: ‘Least-invasive fallback’ can go straight into customer acceptance criteria - far more concrete than ‘we are safe’.
5. Post-train NVIDIA Cosmos 3 Edge for on-device robot control
- Lane: Edge & real-time (on-device inference / control rate) ｜ Evidence:
[O]｜ Source: NVIDIA Technical Blog ｜2026-08 - Key numbers:
4B omni-model·2B Nemotron reasoner·runs on Jetson Thor·receding-horizon control loop - Signals:
closed-loop success reported·open code or weights - Why it matters: The first reproducible path for a world model on-device: post-train, serve on Thor, then evaluate closed-loop, with code and checkpoint released.
6. RD-VLA: implicit test-time compute scaling via latent iterative reasoning
- Lane: Edge & real-time (on-device inference / control rate) ｜ Evidence:
[R]｜ Source: arXiv 2602.07845 ｜2026 - Key numbers:
constant memory regardless of reasoning depth·up to 80x inference speedup vs token-based reasoning VLAs - Signals:
latency or control-rate numbers - Why it matters: Robots can ‘think longer’ without paying token latency - the key to moving test-time compute inside the control loop.
Lane distribution (picks, last 7 days)
| Lane | # |
|---|---|
| Foundation models (VLA / WAM) | 2 |
| Data engine (collection / curation / scaling) | 4 |
| Training & self-improvement (RL / post-training) | 6 |
| Simulation & evaluation (sim2real / closed-loop) | 12 |
| Edge & real-time (on-device inference / control rate) | 6 |
| Safety, permissions & compliance | 1 |
| Systems & orchestration (long-horizon / multi-policy) | 5 |
| Embodiment & supply chain (hardware / cost) | 8 |
Recent cadence
- Source mix (picks, last 7 days):
[O]14 ·[R]17 ·[M]13 - Picks per recent run: 09-22 8 · 09-23 6 · 09-24 2 · 09-25 8 · 09-26 8 · 09-27 8 · 09-28 4
Standards & compliance watch
- ISO 25785-1 — Safety requirements for dynamically stable industrial mobile robots — Under development; committee-stage as of 2026-09.
- Regulation (EU) 2023/1230 (Machinery Regulation) — Fully applicable 2027-01-20.
- EU AI Act official timeline (Council of the EU) — High-risk deadlines 2027-12-02 / 2028-08-02 per Digital Omnibus.
Per-item analysis on this page was drafted by radar-analyst (claude-opus-5.5) and merged automatically after deterministic checks and an agent review, without human review; titles and numbers are untouched.
This repository is technical observation. It is not investment advice, a product commitment, or a safety certification conclusion. Paper claims are author-reported.