MEKAWIRE
News

Wire

Every funding round, launch and paper, as it's reported.

21 stories
AllBreakingAnalysisFundingM&AResearchProductPolicyGlobalUSCNKRSG
6 days ago
Research—Single source

AWS publishes an open-source Physical AI Toolchain built around NVIDIA Isaac, GR00T and Cosmos

Amazon Web Services introduced the Physical AI Toolchain on AWS in a blog post on Oct 7, a public set of reference architectures, infrastructure-as-code templates and deployment automation for building industrial robots, autonomous mobility and humanoids; Amazon described it again in a company post on Oct 8. The pipeline covers synthetic data generation, training, simulation and validation, edge deployment and continuous improvement, pairing SageMaker, EC2 GPU instances, S3 and IoT Greengrass with NVIDIA Isaac Sim, Isaac Lab, Isaac GR00T, Cosmos and OSMO. The code sits in the aws-samples GitHub organisation under the Apache 2.0 license, and customers can use the whole stack or single pieces. It is the second open physical-AI stack tied to NVIDIA in two weeks that MEKAWIRE has covered, after NVIDIA's open agent safety platform on Sep 28.

1 sourceSynthetic Data Generation
6 days ago
ResearchCH

ETH Zurich turns an off-the-shelf robot hand into a robot that walks on its fingertips

Researchers at ETH Zurich's Soft Robotics Lab describe in an arXiv preprint (2609.17172) a commercial Wuji Technology robot hand fitted with an 80-gram backpack holding a battery, an IMU and a Raspberry Pi Zero, which crawls on its fingertips with no arm attached, IEEE Spectrum reported on Oct 8. The controller was trained with reinforcement learning in simulation, using virtual springs to hold a crawling posture; the hand walked on 14 surfaces at about 9 cm per second, turned both ways, righted itself in 21 of 25 trials and pushed a 41-gram cube to targets up to 40 cm away in 15 consecutive trials. Lead author Amirhossein Kazemipour is a Ph.D. candidate in the lab. The work is a preprint and has not been peer reviewed.

1 source
14 days ago
ResearchUS

Anthropic finds robots can do 74% of US physical tasks but are cost-competitive on just 0.3% of work

Anthropic economists scored about 19,000 O*NET job tasks and concluded that robots can technically perform 74% of physical tasks in the US, yet beat human labour on cost for only 0.3% of tasks. At the historical 3% annual decline in robot prices, the study estimates it would take 40 years for that share to reach 10%.

2 sourcesAnthropicFigure AISkild AIDexterous ManipulationRobot Foundation Models
14 days ago
Research—Single source

Innodata builds Vicon-equipped motion-capture lab to supply humanoid training data

Data engineering company Innodata Inc. has opened a motion-capture laboratory in New Jersey to gather real-world movement data for training humanoid robots and other physical AI systems, The Robot Report reported on Sept. 30. The lab uses Vicon infrared optical tracking cameras that record the 3D motion of human and robot bodies with sub-millimeter accuracy and millisecond latency. Innodata said it will sell ready-made motion datasets, run custom collection projects, retarget data across robot platforms and offer independent performance and safety validation of robots.

1 source
26 days ago
ResearchUS

Video Friday: Two Birotors Make a Quadrotor

Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months.

1 source
Aug 20, 2026
ResearchCN

Tsinghua's vision-driven soccer humanoid makes the cover of Science Robotics, trained entirely in simulation

A Tsinghua University team working with ByteDance Seed and China Agricultural University published a reinforcement learning controller in Science Robotics on August 19, 2026 that links a humanoid's camera directly to its legs. On Booster Robotics' T1, the authors report a 46% cut in ball-position error, up to 64% faster time-to-kick and about 90% kick success in frontfield positions, with no real-world fine-tuning.

4 sourcesBooster RoboticsUnitree RoboticsNVIDIARLSim-to-Real Transfer
Aug 20, 2026
ResearchUS

Generalist's GEN-1.5 learns robot tasks from a single 3 to 12 second demonstration, reaching 59% without fine-tuning

Generalist AI said on August 19, 2026 that GEN-1.5 can learn a new manipulation task from one demonstration of 3 to 12 seconds placed in its context, averaging 59% success over 10 tasks with no fine-tuning and 83% after 10 gradient steps on about 5 minutes of data. The company says the ability emerged from more than eight months of pretraining, but calls the tasks simple and the skills brittle, and has named no customers.

1 sourceGeneralist AISkild AIPhysical IntelligenceRobot Foundation ModelsImitation Learning
Jul 17, 2026
ResearchUS

MIT's NeuralActuator lets low-cost robot arms feel force without sensors, winning RSS 2026's systems award

NeuralActuator, a model from MIT's Computational Design and Fabrication Group with Amazon, won the Outstanding Systems Paper Award at RSS 2026. It learns how cheap servo motors really behave from motor current, voltage and temperature, and estimates external force with a reported mean error of 0.12 N, against 0.66 N to 1.41 N for classical baselines.

4 sourcesElectric ActuatorsSim-to-Real Transfer
Jul 17, 2026
ResearchKR

RSS 2026's top paper, FlashSAC, cuts humanoid walking training on one GPU from hours to minutes

FlashSAC, an off-policy reinforcement learning method from a team spanning Holiday Robotics, KAIST, KRAFTON, TU Darmstadt, KTH and others, won the Outstanding Paper Award at Robotics: Science and Systems 2026 in Sydney. The authors report training a Unitree G1 walking policy in about 20 minutes on a single RTX 5090, against about three hours for the standard PPO method.

5 sourcesUnitree RoboticsNVIDIARLSim-to-Real Transfer
Jun 5, 2026
ResearchUS

Amazon FAR's OmniRetarget takes ICRA 2026's top paper prize for turning human motion into humanoid training data

OmniRetarget, a motion retargeting method from Amazon's Frontier AI & Robotics team and university collaborators, won both the Best Conference Paper Award and the Best Paper Award on Robot Manipulation and Locomotion at IEEE ICRA 2026 in Vienna. The method preserves contacts between body, objects and terrain, and the authors used it to generate more than 8 hours of trajectories that trained a Unitree G1 to perform parkour and box carrying in the real world without real-robot fine-tuning.

7 sourcesAmazonUnitree RoboticsBoston DynamicsRLSim-to-Real Transfer
May 18, 2026
ResearchKR

Korea launches 50.4 billion won humanoid programme with KIST and three LG affiliates, putting solid-state batteries in robots

Korea's science ministry on May 18 started a five-year, 50.4 billion won programme to build a national AI humanoid platform led by KIST, with LG Electronics, LG AI Research and LG Energy Solution among the partners. The plan includes a vision, touch, language and action model, an all-solid-state battery for humanoids and a field test of at least 20 robots in care settings.

2 sourcesLG ElectronicsVLARobot Foundation Models
Apr 23, 2026
ResearchJP

Sony AI's Ace becomes the first robot to beat elite table tennis players under official rules, Nature paper shows

Sony AI's autonomous table tennis system Ace won three of five matches against elite players and lost both matches against two Japanese T.League professionals in April 2025 trials, according to a paper published in Nature on April 22, 2026. The robot combines event-based vision sensors from Sony Semiconductor Solutions, an eight-axis arm and reinforcement learning policies trained in simulation, and reaches a perception latency of 10.2 milliseconds.

2 sourcesSonyFANUCYaskawa ElectricRLSim-to-Real Transfer
Apr 16, 2026
ResearchUS

Toyota Research Institute's blind trials in Science Robotics find multitask robot models learn new tasks with a fraction of the data

An 82-author team from Toyota Research Institute, Cornell University and MIT reported in Science Robotics on April 15, 2026 that multitask large behavior models, built on diffusion policy, outperform single-task baselines in blind, randomized trials and need far less task-specific data. The preprint version describes about 1,700 hours of robot demonstrations across more than 500 tasks and 1,800 controlled real-world trials.

3 sourcesToyota Research InstituteBoston DynamicsPhysical IntelligenceImitation LearningRobot Foundation Models
Apr 3, 2026
ResearchUS

Generalist's GEN-1 robot model reaches 99% success on trained tasks with about an hour of robot data, pretrained without any robot data

Generalist AI released GEN-1 on April 2, 2026, an embodied foundation model pretrained on more than 500,000 hours of human physical interaction data captured with wearable devices and no robot data. The company reports 99% average success on tasks where its previous model reached 64%, about one hour of robot data per task and task times roughly three times faster than the prior state of the art; the results are the company's own and have not been independently replicated.

4 sourcesGeneralist AIPhysical IntelligenceSkild AIRobot Foundation ModelsImitation Learning
Jan 28, 2026
ResearchUS

Figure's Helix 02 hands a humanoid's whole body to one neural network, deleting 109,504 lines of control code

Figure AI on Jan. 27, 2026 released Helix 02, a robot foundation model that runs a humanoid from camera pixels and fingertip touch down to joint commands at 1 kHz, replacing 109,504 lines of hand-written C++ with a 10-million-parameter learned controller. The company showed a Figure 03 unloading and reloading a dishwasher across a full kitchen in a four-minute, 61-action run it says had no resets and no teleoperation.

6 sourcesFigure AIPhysical IntelligenceNVIDIARobot Foundation ModelsVLA
Nov 5, 2025
ResearchUS

Generalist AI's GEN-0 claims robotics' first scaling laws, trained on 270,000 hours of real-world manipulation data

Generalist AI introduced GEN-0 on November 4, 2025, a family of embodied foundation models that the company says improves predictably with more physical interaction data. It reported a capacity threshold at 7 billion parameters, models scaled past 10 billion, and a pretraining corpus of more than 270,000 hours growing by over 10,000 hours a week. The findings were published as a company blog post, not a peer-reviewed paper.

3 sourcesGeneralist AIPhysical IntelligenceNVIDIARobot Foundation ModelsDexterous Manipulation
Sep 26, 2025
ResearchGB

Google DeepMind's Gemini Robotics 1.5 makes robots think before they act, and puts its planning model in developers' hands

Google DeepMind on 25 September 2025 introduced Gemini Robotics 1.5, a vision-language-action model that reasons in natural language before moving and transfers skills across robot bodies, alongside Gemini Robotics-ER 1.5, an embodied reasoning model opened to developers as a preview in the Gemini API. In its technical report, the pairing of the two cut total subtask failures on long-horizon tasks to 22% from 44.5% when a general Gemini 2.5 Flash model did the planning.

4 sourcesGoogle DeepMindGoogleApptronikVLARobot Foundation Models
Aug 21, 2025
ResearchUS

Boston Dynamics and Toyota Research Institute put one 450-million-parameter behaviour model in charge of Atlas's hands and feet

Boston Dynamics and TRI showed Atlas completing a long workshop task under a single Large Behavior Model that commands walking and manipulation together. The 450M-parameter Diffusion Transformer runs at 30 Hz on camera, body-state and language inputs and learns new behaviours from VR teleoperation data without new code. No success rates or data volumes were disclosed.

4 sourcesBoston DynamicsToyota Research InstituteToyota Motor CorporationRobot Foundation ModelsImitation Learning