Robotics Weekly: Amazon's Million-Robot Fleet, UCLA's PhoneBot, and NeodraDynamics' Hydraulic Humanoid
This week's digest covers Amazon's milestone of 1 million warehouse robots, NeodraDynamics' 13-foot hydraulic humanoid, and UCLA's $400 Android-based PhoneBot. We also report on Johns Hopkins' heart-docking micro-bots, Ukraine's drone-dropped robotic units, and Tesla's new patent for robot skin.
Podcast В· 3 min
Amazon reaches 1 million robot milestone
Amazon has surpassed 1 million robots deployed across its warehouse network, integrating AI-driven traffic control and touch-sensitive robotic arms. The company's new 'DeepFleet' AI system routes robots to reduce travel time by 10%, while its Vulcan arm is designed to handle 75% of item types. Amazon reports that its global incident rate fell 43% from 2019 to 2025, with internal data linking robotics to a 40% reduction in severe injuries, though minor injuries have increased by 77%.
NeodraDynamics unveils 13-foot hydraulic humanoid
Beijing-based NeodraDynamics has introduced Neo1, a 13-foot, 1.2-ton humanoid robot capable of a 1,102-pound payload. Unlike the industry trend toward all-electric actuators, Neo1 utilizes a hydraulic system for 39 of its 41 joints, which the company claims provides superior force. While the robot is being positioned for heavy-duty industrial tasks in ports and mines, it has yet to be deployed in a real-world factory setting.
UCLA turns Android phones into humanoids
Researchers at UCLA have developed 'PhoneBot,' a 19-inch humanoid that uses an Android smartphone as its primary brain. With a hardware cost of approximately $400, the robot uses 13 motors to walk and recover from falls. While currently limited to flat surfaces and lacking arms, the project aims to provide an accessible platform for students and small labs to enter the field of humanoid robotics.
Tesla files patent for touch-sensitive robot skin
Tesla has filed a patent for a flexible, touch-sensitive robot skin designed to fit curved surfaces like fingers and palms. The patent describes a method of printing sensor patterns onto flat materials before molding them into shape. This development comes as the company continues to refine its Optimus humanoid, which has reportedly faced challenges with hand assembly and sensor reliability.
Johns Hopkins builds heart-docking micro-bots
Engineers at Johns Hopkins have developed magnetic micro-bots capable of navigating a beating heart to deliver targeted antibiotic doses. The system consists of a barbed needle module that anchors to infected tissue, acting as a docking station for a drug-delivery module. While currently tested only in artificial hearts and pig tissue, the technology offers a potential alternative to long-term IV antibiotics for treating serious heart infections.
Ukraine deploys drone-dropped ground robots
Ukraine is utilizing heavy drones to transport and drop armed ground robots behind Russian lines, a tactic described as 'marsupial' robotics. During Operation Vivaldi, units reportedly dropped explosive-laden machines over 10 km behind enemy positions. Ukraine has ordered 25,000 ground robots for the first half of 2026, with the goal of replacing up to 30% of personnel in high-risk sectors by the end of the year.
EPFL develops thread-like motor
Researchers at EPFL have created 'FiberMotor,' a flexible, thread-like motor that uses electrostatic attraction to generate movement. The motor can be stitched into clothing, offering a potential solution for soft exosuits and haptic wearables without the bulk of traditional gears. The team is currently working on improving the motor's durability and strength for commercial applications.
Robot gyms scale data collection
Robotics companies are increasingly investing in 'robot gyms' to generate the real-world training data needed for physical AI. Figure AI is reportedly paying approximately 44,000 participants to record themselves performing tasks, with plans to spend over $1 billion on data generation over the next year. This trend highlights the industry's shift toward human-demonstration data to improve robot adaptability.