Yango Tech Robotics @YangoRobotics

We’ve just added a new motion primitive to our stack: slide. Sounds trivial, but keeping an object sliding stably needs controlled force, continuous contact, friction modelling, and real-time correction. The robot learned it purely from teleop demos – no hard-coded scripts and it now lives next to pick/place/push in the behaviour library. Why it matters: most home & industrial workflows are just chains of micro-skills like this, so the richer the primitive set, the more useful Physical AI becomes. #robotics #PhysicalAI

We’re mixing open-source data with our own datasets - and all the smart hybrids in between. Goal: make our models learn faster, adapt better, and hit higher precision. Strong foundation model = new skills in less time. Set the right parameters, let the system cook.

The next phase of AI won’t live only in the cloud - it will be *embodied*. Systems that can see, reason and act in the physical world, safely and efficiently. Recent Forbes Tech Council insights line up with what we’re seeing on the ground every day. We’re moving from disembodied models to embodied intelligence: by 2028, 60%+ of industrial AI systems are projected to include physical actuation. Spatial understanding (3D vision, sensor fusion, SLAM) within ±5 mm, low-latency edge infrastructure (those extra 50 ms really hurt), and cross-disciplinary teams that speak both kinematics and ML are becoming non-negotiable. And none of this ships without trust: Physical AI that works in shared spaces has to live inside standards like ISO 10218, ISO 3691-4 and IEC 61508, not just cool demo videos. We’re building for exactly this shift - from simulated learning to embodied execution, bridging foundation models and real-world robotics forbes.com/councils/forbe…

From early prototypes to production-ready systems, our work has always been grounded in real-world data — including some of the very benchmarks first inspired by STIQ Ltd reports years ago. That’s why being included in the STIQ Ltd 2025 Picking Automation Report is such a milestone for us - a recognition of our progress toward building Physical AI that enables robots to understand, adapt, and act in the real world. Huge kudos to the team! Let's keep rocking!

Voice → coffee. UR10 + @Robotiq_Inc obeys. Physical AI isn’t sci-fi, it’s brewing now. ☕️ #PhysicalAI

@SoftBank just bought a seat at the Physical AI table. Software that moves things is where the next platforms get built cnbc.com/2025/10/08/sof… #physicalai

Alibaba + NVIDIA doubling down on Physical AI is exactly the kind of infrastructure play robotics needs. Cloud scale + edge intelligence = game changer🧠 reuters.com/world/china/al…

@AlibabaGroup + @NVIDIAAI doubling down on Physical AI is exactly the kind of infrastructure play robotics needs. Cloud scale + edge intelligence = game changer🧠 reuters.com/world/china/al…

Say it → it happens. We turn voice into structured intents and execute them with guardrails + confirmations when needed. Clear steps today; more free-form next #PhysicalAI

5. Data & safety as first-class citizens. Clean pipelines, robustness metrics, failure taxonomies, recovery policies—reported like KPIs, not appendices

4. World models over sim comfort. Learned dynamics for look-ahead planning and failure prediction; production data > domain-randomization

3. Whole-body control goes practical. Learned + MPC/impedance layers coordinating legs–torso–hands for contact-rich tasks, not stage poses

2. Cross-embodiment transfer. Train on arms, deploy on humanoids with light finetune; reusable motor primitives are becoming standard #CoRL2025 #Humanoids #PhysicalAI

We’re at hashtag#CoRL2025 in Seoul 🇰🇷 Five signals that actually matter: 1. VLA brains, grounded. End-to-end Vision-Language-Action stacks that plan, act, and recover. Less prompt cosplay, more closed-loop control

Locomotion = solved baseline. Now the challenge is intelligence + manipulation: task understanding, context, adaptation, repeatable execution. Our lane: brains & skills - vision + touch, force control, generalizable motion. One head, many bodies.

🇦🇪Local compute wins - @nvidia x UAE: New AI+robotics lab in Abu Dhabi → faster protos closer to the floor lnkd.in/eNzW_NWt

Impressive hand by Pan Motor. 20 active DOFs (4 per finger), weighs <600g, 15N fingertip force, 10kg static grasp, >300,000-cycle lifespan, and 30 arcminute accuracy. youtube.com/watch?v=LXVV-o…

First controlled test: a model trained to make coffee. Not scripted code, but a learned policy. Next: expand data across machines, objects & failure modes → robustness via generalization. Coffee prep is a proxy for Physical AI: perception → action → adaptation. #PhysicalAI

Walmart’s AI “super-agents” prove AI at scale delivers - 40% faster support, shift planning 90→30 min, fashion cycles 18 weeks shorter. That same spirit drives #PhysicalAI: machines that see, adapt & act beyond one use case. lnkd.in/diSjqtxM

Robots with real touch? 👀 Tohoku Uni’s “TactileAloha” fuses sight + touch so bots don’t just see: they feel and adapt. From automation → augmentation techxplore.com/news/2025-09-p…