Robotics

Today's Stories

1. 1

   Price to sales forward of Micropolis AI Robotics – AMEX:MCRP

   Price to sales forward of Micropolis AI Robotics – AMEX:MCRP

2. 2

   Mistral Robotics model puts France's industrial AI push in focus

   Mistral Robotics model puts France's industrial AI push in focus

3. 3

   NVIDIA, Hugging Face Partner to Add Robotics AI to Open-Source Library

   NVIDIA and Hugging Face announced on July 6 a collaboration to integrate NVIDIA's physical AI capabilities into LeRobot, Hugging Face's open-source robotics library. The partnership brings the NVIDIA Isaac GR00T 1.7 reasoning model and the Isaac Teleop framework directly into LeRobot. The move makes NVIDIA's robotics AI tools more accessible to developers working with open-source software. By embedding these capabilities into an existing library, the partnership lowers barriers for teams building and deploying robotic systems without requiring separate proprietary platforms.

   LeRobot is now a more complete robotics development environment. Developers can access the integrated tools through Hugging Face's platform, which already hosts open-source AI models and datasets.

4. 4

   LingBot-Video: open-source video AI with robot action planning

   Researchers released LingBot-Video, an open-source video generation model built as a sparse mixture-of-experts transformer (13B total parameters, 1.4B active). The model includes six-reward reinforcement learning post-training, with one reward focused on physical plausibility, and can generate video from action and hand-pose conditions to predict robot movements. Weights, code, and software stack are openly available. The model ranks at the top average on RBench (a video-generation benchmark), though reasoning-heavy tasks still favor closed models. Its action-to-video capability frames it as a potential world model for robotics planning, bridging video generation and embodied AI—areas where most leading systems remain proprietary.

   The physical-plausibility reward is graded by a vision-language model (VLM) sampling frames, which researchers partially address by adding real-video negatives to prevent reward hacking. The model is second on general text-to-video tasks; whether VLM-judged physics translates to closed-loop robot performance remains unanswered, and no robot execution numbers are published yet.

5. 5

   Waymo's new robotaxi stuck in free-ride limbo by California regulator delay

   Waymo's application to the California Public Utilities Commission to expand its service area and add its new Ojai vehicles to its fleet remains pending. The company cannot yet charge passengers for rides in the Ojai, a pale blue Chinese-made car that started picking up riders last month, while it continues to charge for rides in its Jaguar I-PACE robotaxis. Unlike other states that allow robotaxis to launch with minimal oversight, California requires approval from both the Department of Motor Vehicles and the Public Utilities Commission before companies can carry paying passengers. This regulatory requirement is delaying Waymo's expansion into Northern California (from Sea Ranch and Sacramento through Berkeley, Oakland, and San Jose) and Southern California (from Los Angeles into Thousand Oaks, Santa Clarita, and down to the Tijuana border past San Diego).

   Waymo's Ojai rides could remain free until the end of September and beyond if the regulatory delay persists, creating an unusual situation where one vehicle type in the fleet is subsidized while another generates revenue.

6. 6

   AI-powered robot automates Serrano ham labeling, first in meat industry

   Spanish automation firm Timpolot combined an AI vision system with a Stäubli SCARA robot to label Serrano hams automatically, processing 150,000 to 180,000 kg per day at up to 900 pieces per hour. The system uses AI to identify each ham's bone position and determine the optimal labeling point, then injects labels via a pneumatic fastener applicator guided by camera coordinates. Serrano ham labeling was previously only possible by hand because bones account for 30 to 40 percent of each ham's weight and vary in position unpredictably; human operators had to avoid bone areas to prevent tool damage, creating physical strain and requiring expertise. Automation removes this bottleneck, allows workers to move to higher-value tasks, and centralizes traceability control through unified IT management—addressing a real constraint in a major food production industry.

   The system has been running for several months to the client's complete satisfaction. The robot could operate faster than its current 750 pieces per hour peak of 900, but label printing and upstream processes limit overall speed. The Stäubli TS2-80 HE was selected for food-grade lubrication, hygiene-rated surfaces, and long-term mechanical reliability.