The Federal Trade Commission has issued a sweeping ban on foreign imports of advanced robots, including humanoids, quadrupeds, and wheeled robots, citing national security concerns and the need to protect domestic robotics companies from Chinese competition. The decision aims to align political and industry interests and follows the Trump administration's aggressive approach to protecting the US AI industry.
China has threatened retaliation against US restrictions on foreign-produced humanoid robots, warning that the measures will damage economic and trade relations between the two countries. The US Federal Communications Commission added foreign-made advanced robotic devices to its Covered List, preventing new models from receiving necessary authorizations for import, marketing, and sale in the United States. China's Ministry of Commerce accused the FCC of ignoring Beijing's approach to product restrictions, warning that it would introduce countermeasures if the decision remained in place.
1X's Norwegian-American robotics company has unveiled its soft and helpful robot companion, Neo, with five-finger hands designed to replicate human hand movement. The hands have 25 degrees of freedom of movement, can grip odd shapes, detect slipping objects, and move extremely quickly. Neo's goal is full automation, but for now, it's partly teleoperated, allowing humans to remotely control the robot.
Walden Robotics, a spinout from Toyota Research Institute (TRI), has emerged with $300 million in funding at a valuation of $1.1 billion, focusing on practical humanoid robots for manufacturing and logistics environments. The company's approach is centered around designing a robot that can safely and efficiently meet specific needs, skipping the legs to focus on higher payload requirements. Walden's partnership with Toyota provides a people-first culture that emphasizes improving quality of life, while also targeting high-utilization applications in manufacturing, making economics viable. The company aims to develop "multi-purpose" robots through leveraging previous learned skills, aiming for a general purpose robot platform.
Figure 03's autonomous ladder climbing capability marks a significant milestone, utilizing advanced "loco-manipulation" and balance systems introduced in its Helix 02 software architecture. CEO Brett Adcock sharply criticizes wheeled robotics as an "utter dead end" due to environmental limitations like stairs and ladders. The robot's current pace is methodical but will eventually match human speed, addressing online critiques of the platform's agility.
protoros2 is a middleware wrapper and orchestration engine designed to provide seamless Protobuf serialization in ROS/ROS 2 without compromising performance. It acts as a non-intrusive "Tri-State Orchestration Engine" that binds third-party Open-Source foundations into a unified architecture, offering features such as multi-channel communication and enterprise security built-in. protoros2 supports multiple "Single Source of Truth" architectures seamlessly and is compatible with the ROS 2 executor ecosystem.
Reimagine Robotics, an AI robotics company founded by former Google DeepMind Applied Robotics team leaders, has emerged with technology that allows robots to learn from workers. The company's robots can be trained and used by anyone, without requiring specialist programmers. Workers can show the robot what to do, correct it on the spot, and move on to the next task, a process called 'monkey-see, monkey-do'.
Researchers have unveiled Milo, the world's first fully autonomous robotic guide dog designed to assist blind and visually impaired people. The AI-powered robot can navigate both indoor and outdoor environments without requiring a prior 3D scan, using onboard artificial intelligence to identify paths and avoid obstacles. Milo can be manufactured for less than $2,000, making it a more accessible alternative to traditional guide dogs that cost tens of thousands of dollars.
Unitree Technology expects first-half revenue of about CNY1.1 billion ($156 million) for 2026, representing year-on-year growth of between 35.6 percent and 45.4 percent. The company plans to raise 4.2 billion yuan through its initial public offering (IPO), valued at more than 50 billion yuan, following regulatory challenges from the US Federal Communications Commission that may restrict future access to the US market.
German robotics company Agile Robots expects to double its revenue this year, reaching approximately €600 million, after generating €300 million in 2025. The company has already signed customer contracts supporting its 2026 revenue forecast and aims to reach profitability within two to three years. Founded in Munich in 2018, Agile Robots has received $1.5 billion in funding from investors led by SoftBank Group and expects a significant portion of new business this year will come through acquired companies.
The UK government has launched a £20 million funding competition to support the development of agricultural robots and automation technologies, with the aim of improving productivity and reducing dependence on seasonal labor. The funding is available to agri-tech businesses working with researchers and farmers to develop robotic systems for planting, monitoring, and harvesting crops. The competition follows a recent £53 million funding package, bringing the total available this year to £123 million, and includes projects in livestock, forestry, and ornamental horticulture.
Researchers from Chiba University developed a control system that helps bird-inspired robots stay stable in windy conditions, overcoming unexpected movements and improving disturbance rejection by 53.1 percent. Flapping-wing robots are susceptible to wind gusts, but the new method accounts for their non-minimum-phase behavior, allowing them to maintain stable flight. The study could advance the development of small autonomous flying systems for inspection, monitoring, and search-and-rescue operations in confined spaces.
Dubai Future Foundation and Oxa launch SHIFFT, a joint venture to deploy self-driving vehicles at ports and airports. The venture will utilize Oxa's autonomous technology to increase productivity, efficiency, and safety for customers, directly contributing to Dubai's goal of doubling foreign trade volume by 2033. SHIFFT aims to accelerate the commercialization of Oxa's Industrial Mobile Autonomy (IMA) technology, leveraging autonomy to unlock new levels of efficiency in major industrial environments.
Commissioning a robotic production cell requires testing across various systems, including robot controls, drives, safety functions, and plant networks. A qualified systems integrator reduces commissioning risk by treating these systems as one integrated system with documented acceptance criteria. This includes defining interfaces before the build, performing factory testing to expose logic and interface problems, and rechecking the cell in its final environment under site conditions.
A Qualcomm-powered humanoid robot named 4NE-1 collapsed spectacularly on stage during a company's keynote at Computex 2026, triggered by a brief communication glitch. The robot performed its "safe-collapse" sequence, lowering itself to its knees, and the safety systems worked as intended. Stagehands quickly cloaked and carried off the stricken robot, which had been powered by Qualcomm's Dragonwing IQ10 robotics reference design.
Boston Robot Hackers is hosting an August meeting with speaker Shivam Chopra, PhD, on Forward and Inverse Kinematics. The event will take place on August 6, 2026, from 7:00pm to 9:00pm at Artisans Asylum in Alston, Boston.
HEBI Robotics has been awarded a Phase I contract from NASA's Small Business Innovation Research program to develop compact robotic actuators for use in small satellites and other space missions. The project aims to design miniature actuators that can generate high levels of force while remaining small enough for spacecraft with limited space and weight. The technology could be used in CubeSats and other compact spacecraft where existing robotic hardware is often too bulky or lacks flexibility.
A planning-control framework is proposed for Minimum-Lap-Time (MLT) drifting in autonomous vehicles. The framework uses a Track-Guided Reinforcement Learning (TgRL) drift control method to learn a drift racing policy that optimizes both vehicle motion control and lap time reduction. The reward function incorporates an instant reward term and an end reward term derived from the MLT objective, allowing the agent to learn progressively from drift control policies to comprehensive drift race policies. Simulation results demonstrate effective learning of a drift racing policy that reduces lap times while maintaining vehicle motion control performance.
A motion planning framework generates dynamically feasible and collision-free trajectories for autonomously opening and traversing both push and pull doors using a nonlinear Model Predictive Control (MPC) optimization framework. The robot and door are formulated as a coupled dynamical system, with manipulation feasibility enforced through a penalty-based constraint. Simulations and a hardware experiment demonstrate the approach's success in planning feasible trajectories for door traversal.
Researchers Fanxing Meng and Jing Xiao developed a method to enable humanoid robots to learn combined manipulation and locomotion skills using interaction representation and skill composition. The approach learns a whole-body reaching and grasping policy via spatial convolution, followed by combining it with a separately learned getting-up policy to create a more complex policy. This results in a 93% zero-shot success rate for grasping unseen objects and a 96-100% success rate for standing up while holding the object.
Researchers Nataliya Nechyporenko, Ava Abderezaei, and Alessandro Roncone introduced a distributed contact model called Compliant Sphere Lattice Contact (CSLC), which enables physically accurate contact modeling for sphere-based robot representations. CSLC models the robot interface as a compliant lattice of surface spheres connected through anchor and lateral springs, improving accuracy by producing a spatially distributed contact patch when pressed against an object.
A localization framework uses spatiotemporal fields governed by partial differential equations (PDEs) as signatures for vehicle localization. The PDE classes considered include the shallow water equations and the advection-diffusion equation, which model free-surface flows in coastal and riverine environments and transport and mixing of scalar quantities. A Rao-Blackwellized particle filter improves localization accuracy by fusing multiple field channels. Simulation studies demonstrate that this framework outperforms a standard particle filter in terms of final position error and Root Mean Square Error (RMSE).
A hybrid attention estimation pipeline was developed for adaptive human-robot interaction (HRI) using an expressive robotic head based on the InMoov ecosystem. The pipeline combines a fast geometric perception layer with an independent semantic perception layer, providing high-frequency face and head-pose information and contextual attention labels. The system was evaluated in 40 trials across baseline and adaptive interaction conditions, showing reliable interaction start and non-redundant semantic information between outputs.
Researchers introduced the problem of serving robot policies to multiple robots from a remote GPU, formulating it as a scheduling problem. A proposed algorithm accounts for heterogeneity in action chunk consumption rates among robots and improves system throughput by up to 18%. The system, Armory, was validated on fleets of simulated and real robots, revealing that conventional batching methods are insufficient for closed-loop robot policy execution.
A robotic system was developed to automate the manufacturing of dielectric elastomer actuators, specifically focusing on reducing process time and removing human operators. Twenty multilayer devices were fabricated with 12 dielectric layers and 11 alternating carbon nanotube electrodes, resulting in a 14.2% reduction in total process time and eliminating human involvement in 56.1% of the process.