Physical AI Radar · 2026-10-08
A daily radar on the Physical AI frontier: eight lanes, trilingual digests, traceable evidence.
- Generated:
2026-10-08 07:30 UTC - Window:
papers 2026-10-04 → 2026-10-08 · posts 2026-09-08 → 2026-10-08 · no repeats within 30 days (UTC) - Evidence tags: O = first-party official · R = paper/preprint · M = media/secondary
Today’s picks
1. Agentic RSR: Real-to-Sim-to-Real through Scene Reconstruction and Execution-Grounded Robot Policies
- Lane: Simulation & evaluation (sim2real / closed-loop) ｜ Evidence:
[R]｜ Source: arXiv 2610.10479 ｜2026-10-07 - Key numbers:
80%(“task success rate reaches 80% of the simulation”) - Signals:
quantitative results·closed-loop success reported·real-robot experiments - Why it matters: Preprint, not yet peer-reviewed; the text gives real-robot results, closed-loop results and figures. Watch how well the simulated result predicts real-robot performance.
A simulation of a real robot workspace must preserve task-relevant interactions, while policies developed in it must operate on observations available to the real robot.
2. RealtimeWAM: How Fast Can I Run My World Action Model?
- Lane: Simulation & evaluation (sim2real / closed-loop) ｜ Evidence:
[R]｜ Source: arXiv 2610.10079 ｜2026-10-07 - Key numbers:
63.09 ms(“latencies are 24.09 and 63.09 ms”) ·82.75%(“Average success rates are 82.75% and 87.41%”) ·87.41%(“rates are 82.75% and 87.41%”) - Signals:
quantitative results·closed-loop success reported·real-robot experiments·latency or control-rate numbers - Why it matters: Preprint, not yet peer-reviewed; the text gives real-robot results, closed-loop results and figures. Watch how well the simulated result predicts real-robot performance.
World Action Models (WAMs) combine visual dynamics modeling with action generation, but their high inference latency limits responsive robot control.
3. RFPO: Rectified Flow Policy Optimization for Embodied Control
- Lane: Edge & real-time (on-device inference / control rate) ｜ Evidence:
[R]｜ Source: arXiv 2610.10453 ｜2026-10-07 - Key numbers:
2.4%(“one-step returns remaining within 2.4% of their corresponding”) ·98.5%(“one-step execution retains 98.5% of the 64-step”) ·4.39 ms(“mean inference latency from 4.39 ms to 0.08 ms”) ·0.08 ms(“from 4.39 ms to 0.08 ms”) - Signals:
quantitative results·real-robot experiments·open code or weights·latency or control-rate numbers - Why it matters: Preprint, not yet peer-reviewed; the text gives no closed-loop results. Watch whether the speed-up holds on the target hardware at control rate.
Flow-based policies provide an expressive framework for continuous robot control, but their iterative ODE integration incurs substantial inference cost.
4. PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation
- Lane: Training & self-improvement (RL / post-training) ｜ Evidence:
[R]｜ Source: arXiv 2610.08784 ｜2026-10-06 - Key numbers:
53.2%(“threshold by up to 53.2% relative to the”) ·95%(“achieves success rates of 95% on Whiteboard Erasing”) ·90%(“on Whiteboard Erasing and 90% on Pipette Liquid”) - Signals:
quantitative results·closed-loop success reported·real-robot experiments - Why it matters: Preprint, not yet peer-reviewed; the text gives real-robot results, closed-loop results and figures. Watch whether the gain holds outside the training tasks.
Pretrained robotic policies can suffer substantial performance degradation under out-of-distribution (OOD) conditions encountered during deployment, motivating post-training through real-world interaction.
5. 研究了1933篇 IROS 论文，我们看到了机器人学的六项新变化
- Lane: Embodiment & supply chain (hardware / cost) ｜ Evidence:
[M]｜ Source: 雷峰网 Leiphone ｜2026-10-08 - Why it matters: Secondary report, so check the primary source before citing; the text gives no figures. Watch price, availability and who is shipping it in volume.
大模型并没有“吃掉”机器人学。9月27日，IROS 2026 将在匹兹堡开幕。
6. 对话如祺出行COO韩锋：世界模型越强，真实数据才是真壁垒
- Lane: Data engine (collection / curation / scaling) ｜ Evidence:
[M]｜ Source: 雷峰网 Leiphone ｜2026-10-08 - Why it matters: Secondary report, so check the primary source before citing; the text gives no figures. Watch whether the data is reusable beyond the authors’ own robot.
过去一年，“物理AI”几乎成了自动驾驶、车企共同争夺的下一张船票。 世界模型、具身大模型、机器人数据……众多公司从不同切入口布局物理AI，试图成为新风口之下的主角，相比之下，如祺出行选择了一条不那么显眼的路径，不自己训练大模型，而是尝试成为模型背后的数据基础设施。
7. IROS 2026大奖揭晓：长期记忆拿下最佳论文，人形机器人打网球、端托盘双双获奖
- Lane: Foundation models (VLA / WAM) ｜ Evidence:
[M]｜ Source: 雷峰网 Leiphone ｜2026-10-08 - Why it matters: Secondary report, so check the primary source before citing; the text gives no figures. Watch whether it cuts the data a new task or embodiment needs.
在机器人领域，一个更大的模型并不能自动解决所有问题。当地时间 9 月 30 日，IROS 2026 主会在美国匹兹堡落下帷幕，本届大会的主要奖项也在最后一天正式揭晓。
8. Harmonic Drive to share how integrated actuators can reduce engineering at RoboBusiness
- Lane: Embodiment & supply chain (hardware / cost) ｜ Evidence:
[M]｜ Source: The Robot Report ｜2026-10-07 - Why it matters: Secondary report, so check the primary source before citing; the text gives no figures. Watch price, availability and who is shipping it in volume.
Attendees will learn how integrated actuators can reduce engineering effort by simplifying system integration.
Frontier baseline (curated)
1. Newton: open-source GPU physics engine for robotics (NVIDIA, Google DeepMind, Disney Research)
- Lane: Simulation & evaluation (sim2real / closed-loop) ｜ Evidence:
[O]｜ Source: NVIDIA Developer ｜2026 - Key numbers:
built on Warp + OpenUSD·MuJoCo Warp up to 152x (locomotion) / 313x (manipulation) vs MJX on RTX 4090 - Signals:
open code or weights - Why it matters: A differentiable new simulation base plus Isaac Lab’s multi-backend switching lets training and evaluation reuse one actuator configuration.
2. π*0.6 / Recap: a VLA that learns from experience via advantage-conditioned policies
- Lane: Training & self-improvement (RL / post-training) ｜ Evidence:
[R]｜ Source: arXiv 2511.14759 ｜2026 - Key numbers:
RL from autonomous rollouts + human corrections·value function fine-tuned on online data - Why it matters: The methodological basis for ‘deployment is training’: to use it, operations must first make intervention logs and success judgements first-class data.
3. XPENG commissions its humanoid production lines; IRON walks off the line autonomously
- Lane: Embodiment & supply chain (hardware / cost) ｜ Evidence:
[O]｜ Source: PR Newswire ｜2026-09-08 - Key numbers:
>80% of core processes automated (reported)·target 1,000+ units/month by year-end (reported) - Why it matters: Line commissioning is a verifiable fact; the monthly target is reported second-hand. Separating the two is the easiest citation mistake to get caught on this cycle.
4. 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.
5. RoboHarness: memory-driven orchestration of heterogeneous robot policies
- Lane: Systems & orchestration (long-horizon / multi-policy) ｜ Evidence:
[R]｜ Source: arXiv 2607.18060 (Huawei Noah’s Ark Lab) ｜2026 - Key numbers:
VLA + RL + TAMP as agentic skills·capability-aware routing·Memory Bridge for policy handoff without retraining - Why it matters: Long-horizon systems are an orchestrator plus a policy pool; the operated object becomes a set of declarable capability boundaries rather than one model.
6. HumanEgo: zero-shot robot learning from minutes of human egocentric video
- Lane: Data engine (collection / curation / scaling) ｜ Evidence:
[R]｜ Source: arXiv 2605.24934 ｜2026 - Key numbers:
30 min/task -> 92.5% average success on 4 real tasks·15 min -> 75%·+41% over matched-time teleoperation - Why it matters: It pushes the cost per new task down to minutes of data - the most direct yardstick for whether your data strategy is behind.
Lane distribution (picks, last 7 days)
| Lane | # |
|---|---|
| Foundation models (VLA / WAM) | 11 |
| Data engine (collection / curation / scaling) | 3 |
| Training & self-improvement (RL / post-training) | 5 |
| Simulation & evaluation (sim2real / closed-loop) | 13 |
| Edge & real-time (on-device inference / control rate) | 9 |
| Safety, permissions & compliance | 4 |
| Systems & orchestration (long-horizon / multi-policy) | 4 |
| Embodiment & supply chain (hardware / cost) | 7 |
Recent cadence
- Source mix (picks, last 7 days):
[O]1 ·[R]40 ·[M]15 - Picks per recent run: 10-02 8 · 10-03 8 · 10-04 8 · 10-05 8 · 10-06 8 · 10-07 8 · 10-08 8
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.
This repository is technical observation. It is not investment advice, a product commitment, or a safety certification conclusion. Paper claims are author-reported.