Robotics news and engineering summaries

Hyundai is acquiring SoftBank's remaining 9.65% stake in Boston Dynamics for $325 million, gaining full ownership to integrate Atlas humanoid robots into its Georgia manufacturing plants by 2028. This move moves beyond demo videos, focusing on long-term factory reliability and vertical integration, while SoftBank shifts its focus toward large-scale AI infrastructure.

Microduck is a $399, 25 cm open-source bipedal robot weighing 800 g. It ships with seven reinforcement-learning behaviors, trained in MuJoCo simulation and deployable on the real robot. Its SDK, simulator, and RL training stack are available on GitHub under Apache-2.0, enabling users to retrain, refine, and share policies. Pre-orders ship before Christmas 2026.

This project presents a compelling case study challenging the notion that AI cannot influence the physical world. By utilizing Claude Code as a virtual farm manager, the team successfully managed a corn crop from seed to harvest. Instead of operating machinery directly, the AI functioned as a central brain, orchestrating complex systems, analyzing IoT sensor data, and coordinating human contractors. Claude Code processed weather forecasts and satellite imagery to make critical decisions regarding irrigation and planting. This experiment demonstrates that AI can effectively bridge the gap between digital processing and tangible agricultural production through strategic coordination, data aggregation, and real-time decision-making, providing a documented blueprint for future autonomous resource management.

A Waymo robotaxi was involved in a collision with a child near a Santa Monica elementary school during morning drop-off hours. The National Highway Traffic Safety Administration (NHTSA) has launched an investigation into the incident, which resulted in minor injuries to the pedestrian. Waymo reported that the vehicle was traveling at 17 mph and decelerated to 6 mph after the child emerged from behind a tall SUV. This incident adds to existing federal scrutiny regarding Waymo autonomous vehicles, including probes into illegal school bus passing. The investigation focuses on whether the AI system exercised appropriate caution in a school zone environment characterized by high pedestrian density and double-parked vehicles.

CARA 2.0 is an upgraded, low-cost quadrupedal robot project designed as a senior capstone. By utilizing re-wound, budget-friendly BLDC motors and custom firmware for affordable FOC controllers, the team achieved high dynamic performance while reducing costs by half compared to the 1.0 version. The project demonstrates advanced mechanical balancing, gait optimization, and effective 3D-printed structural design.

OOMWOO is an open-source, build-it-yourself robot vacuum project designed for makers. It uses Raspberry Pi, ROS 2, and 2D LiDAR for autonomous navigation without requiring cloud services. The project emphasizes local-first control, 3D-printable parts, and a modular architecture, enabling community-driven development through a transparent, build-in-public approach for full digital independence.

A creative hobbyist transformed a spare 42u server rack into a hydroponic farm using a flood and drain system. Utilizing rock-wool, clay balls, and LED lights, the setup is controlled via cronjobs on a PDU. Despite initial challenges with algae and floating pots, the project successfully produced lettuce and herbs inside a server cabinet.

A developer successfully trained a reinforcement learning policy for a drone to survive motor failures by addressing action saturation and reward shaping. By implementing a tanh squashing function and optimizing survival rewards, the agent learned to handle single, dual, and some triple motor failures within a simulation, demonstrating robust generalization.

CERN engineers have developed high-speed levitating karts to replace bicycles for transport within the Large Hadron Collider tunnel during the Long Shutdown 3 maintenance period. Powered by superconducting engines utilizing the Meissner effect, these vehicles may also have future aerospace applications. The project was inspired by designs from local nursery school children.

This project showcases a low-cost guided rocket system prototype built for $96 using 3D-printed parts and consumer electronics. Featuring an ESP32 flight computer, MPU6050 IMU, and canard stabilization, it integrates GPS and telemetry. The repository includes Fusion 360 CAD files, firmware, and OpenRocket simulations, detailing the full mechanical and electronic engineering process.

MicroVision has announced plans for a solid-state automotive LiDAR sensor priced under $200, aiming for a long-term goal of $100. This drastic price reduction seeks to move LiDAR from high-end autonomous vehicle programs into mainstream advanced driver-assistance systems (ADAS) by prioritizing manufacturability and high-volume production over maximum range or resolution.

FreeCAD 1.0 is a powerful, open-source parametric 3D modeler for professional design, mechanical engineering, and architecture. It allows users to convert 2D sketches into 3D objects, supports various file formats, and features tools for FEA, BIM, and robot simulation without licensing fees or vendor lock-in.

Waymo has paused robotaxi services in Atlanta and San Antonio due to vehicles struggling with heavy rain and flooding. Despite a recent software recall intended to mitigate these risks, the autonomous systems failed to navigate flooded conditions effectively, prompting further scrutiny from the NHTSA and NTSB regarding safety and incident reporting.

Tesla's Robotaxi fleet in Austin reported five new crashes via NHTSA data, bringing the total to 14. Analysis shows a crash rate four times higher than human drivers. Tesla remains the only operator fully redacting incident narratives and recently upgraded a past report to include a hospitalization previously undisclosed to the public.

Robotics startups like Figure AI, Sunday Robotics, and Weave Robotics are focusing on laundry as an ideal, low-stakes task to enter the home market. While challenging due to deformable objects, laundry serves as a gateway to gathering data for general-purpose AI. Despite high costs and previous industry failures, companies hope rapid technological advancements will finally deliver functional home robots.

The comma four is an advanced hardware device designed to bring AI-powered driving capabilities to existing vehicles. By running openpilot, an open-source driver assistance system, it provides features such as lane centering, adaptive cruise control, and automated lane changes. The device features 360-degree vision and supports over 325 car models from 27 major brands, including Toyota, Hyundai, and Ford. With its plug-and-play installation and over-the-air (OTA) updates, comma four offers a seamless way to upgrade standard cars with sophisticated autonomy and dashcam recording functionality.

NASA is upgrading astronaut hygiene for the Artemis II mission with the Universal Waste Management System (UWMS). Developed by Collins Aerospace, this 3D-printed titanium toilet addresses the failures of Apollo-era systems by offering a ergonomic, unisex design with simultaneous waste processing and improved privacy, setting a new standard for future moon and Mars exploration missions.

A developer created a Python-based cycloidal gearbox generator for Fusion 360, designed to produce custom gearboxes for NEMA 17 stepper motors. The project highlights the challenges of 3D printing tolerances and provides a functional Version 3 design achieving a 1:9 gear ratio and 66% efficiency, utilizing PLA material and specific hardware constraints.

Waymo has launched its fully autonomous ride-hailing service for all users in Dallas. Following a successful pilot, the service now supports daily commuting and local errands. Expansion plans include service to Dallas Love Field Airport and upcoming autonomous testing on city freeways to improve transportation accessibility for all residents, including those with medical driving limitations.

Unlocked is a groundbreaking open-source project featuring a lock-picking robot designed to challenge the vulnerabilities of physical security systems and skeleton keys like the TSA 007. Built using an Arduino-based OpenRB-150 board and high-precision Dynamixel servo motors, the device employs a brute-force methodology by feeding copper wire through a hollow key to manually actuate each pin. While traditional lock-picking relies on tactile feedback, this robot systematically tests every possible bitting combination, making it resilient against security pins but limited by entropy. A four-pin lock with four cut depths can be solved in approximately three minutes, whereas increasing the pin count exponentially grows the time required. The project utilizes a combination of FDM, SLA, and DMLS 3D printing for its mechanical components and is licensed under GPL 3.0 to promote community-driven innovation in locksmithing and security research.