Humanoid Factories, Warehouse Swarms, and Robots That Can Finally See Glass

Charm · April 2, 2026 · 2 min read · 5 sources

News

Google DeepMind's Gemini Robotics Models Are Going Into Agile Robots' Industrial Humanoid

Agile Robots is embedding DeepMind's Gemini Robotics models into its Agile ONE humanoid platform, targeting real factory floors. The 'AI flywheel' model here is the key mechanic: more deployments generate training data, which improves the models, which enables more deployments. With 20,000+ installations already in the field, Agile has real data leverage that most humanoid startups don't. modernmechanics24.com

Toyota's Swarm Automation Transport Brings Coordinated AGV Fleets to Warehouses

Toyota Material Handling Europe launched a full AGV system combining its SAI125CB stacker with T-ONE software, capable of running mixed human and automated fleets in the same space. The gradual adoption model is smart product strategy: let warehouses start hybrid and scale up, lowering the barrier to entry for automation buyers. roboticsandautomationnews.com

Research

HEAPGrasp Lets Robots Reliably Pick Up Transparent and Reflective Objects Using Only RGB Cameras

Japanese researchers at Tokyo University of Science cracked one of the stubborn bottlenecks in robotic grasping: transparent and shiny objects that defeat standard 3D vision. HEAPGrasp uses silhouettes instead of optical properties, meaning no special sensors required, just a regular RGB camera. This has immediate implications for food, logistics, and glass-heavy manufacturing lines. roboticsandautomationnews.com

IIT Gandhinagar's Virtual Actuation Space Framework Cuts the Complexity of Controlling Flexible Robots

Tendon-driven continuum robots are promising for surgery and confined spaces, but notoriously hard to control in real time. The VAS framework reduces motion control to just two parameters per section, achieving under 1% error margin, which is the kind of precision needed before these things go anywhere near a human body. interestingengineering.com

Semi-Autonomous Legged Robot Tested for Multi-Target Prospecting on Moon and Mars Analog Terrain

ESA researchers tested ANYmal, a quadruped, running semi-autonomous scientific measurements across multiple rock targets without constant human oversight, a direct response to the communication delay problem on Mars. The result: a robot that walks to targets, measures, and moves on without waiting for Earth, which is the only architecture that makes sense given 4 to 22 minute signal lag. phys.org

Stay Ahead

Delivered each morning.