Robotics Weekly: DoorDash's Drone, Dyna's Laundry Robot, and MIT's Muscle-Powered Swimmer
This week's digest covers DoorDash's new in-house delivery drone, Dyna Robotics' laundry-folding system, ETH Zurich's crawling robotic hand, and MIT's muscle-powered soft robot. We also report on Figure AI's industrial recycling, Tesla's production bottlenecks, and the Pentagon's new Autonomous Warfare Command.
Podcast В· 2 min
DoorDash Unveils In-House Delivery Drone
DoorDash has unveiled its in-house delivery drone, a six-rotor aircraft designed to streamline the kitchen-to-backyard delivery process. The system includes custom loading mechanisms, packaging, and dispatch software that coordinates drones with ground couriers and robots, aiming for delivery times under five minutes. This move follows the company's previous experiments with human-loaded robots and represents a strategic shift toward controlling the entire delivery stack to improve efficiency and reduce costs. By managing the hardware and the software, DoorDash aims to solve the last-mile bottleneck that has challenged competitors like Wing and Zipline. This development signals a broader industry trend where delivery platforms are increasingly vertically integrating their hardware to achieve the scale required for profitable autonomous operations.
Dyna Robotics Introduces Taku Laundry Robot
Redwood City-based startup Dyna Robotics has introduced Taku, a wheeled, two-armed robot designed to automate the entire laundry workflow, from loading machines to folding and stocking shelves. The robot features a folding body and steerable wheels, allowing it to reach deep into dryers and navigate tight spaces, while its Dyna-2.1 system manages complex tasks like interrupting folding to handle a finished dryer load. This development is significant as laundry has become a key proving ground for physical AI, with companies like Figure and Physical Intelligence also working on fabric manipulation. Dyna's focus on mastering the entire workflow aims to reduce the need for human supervision in commercial environments like hotels, potentially unlocking significant labor savings.
Figure AI Demonstrates Industrial Recycling
Figure AI has released footage of its F.02 humanoids autonomously leaping into a Finnish arc furnace, with the recovered metal being repurposed into limited-edition collectibles. This demonstration serves as a unique showcase of the robot's capabilities in industrial environments, specifically in recycling and material handling. While primarily a branding and recycling initiative, it underscores the growing trend of deploying humanoids in heavy industrial settings. The event highlights the company's focus on demonstrating real-world utility beyond basic manipulation, aiming to build public and industry confidence in humanoid reliability.
Tesla Faces Optimus Production Bottlenecks
Tesla is reportedly facing production bottlenecks with its Optimus humanoid program, specifically regarding hand manufacturing and supplier constraints. Despite claims of producing hundreds of units weekly, these issues are complicating the company's push toward mass production. The situation reflects the broader challenges of scaling humanoid manufacturing, where complex components like dexterous hands remain a significant hurdle. Tesla's ability to overcome these supply chain and engineering difficulties will be critical to its goal of deploying Optimus at scale, a key pillar of its long-term AI strategy.
ETH Zurich Teaches Disembodied Hand to Crawl
Researchers at ETH Zurich have developed a method to teach a commercial robotic hand to crawl, balance, and operate a keyboard without an arm attached. By mounting a Raspberry Pi, motion sensors, and a battery onto a WUJI hand, the team used reinforcement learning to coordinate the fingers for locomotion across various surfaces, including grass and gravel. This project builds on previous research into detachable robotic limbs and demonstrates how learned control can grant new mobility to disembodied parts. The ability to send individual robotic components into spaces inaccessible to full-bodied robots could have major implications for search-and-rescue and industrial inspection tasks.
MIT Develops Muscle-Powered Soft Robot
MIT engineers have created a wafer-thin, muscle-powered robot capable of swimming through water and navigating mazes. The device uses genetically engineered mouse muscle cells that contract when exposed to blue light, allowing researchers to steer the robot by pulsing light at its fins. This bio-hybrid approach produced 20 times more force per unit of tissue volume than previous designs and remained functional for over 30 days. While currently a lab prototype, the research highlights the potential for soft, flexible robots that are cheaper to build and better suited for delicate work in fragile aquatic environments.
Pentagon Launches Autonomous Warfare Command
The U.S. Pentagon has announced plans to launch the Autonomous Warfare Command, a new military entity dedicated to scaling the development and deployment of AI-driven drones and robots. This move marks a significant step in the integration of autonomous systems into military operations, aiming to streamline the transition from research to field deployment. The command will focus on coordinating the rapidly expanding ecosystem of military robotics, reflecting the strategic importance of autonomous technology in modern defense. This development is likely to accelerate the adoption of robotic systems across various branches of the armed forces.