A US startup, Kall Morris Inc, has successfully tested its Responsive Engaging Arms for Captive Care and Handling (REACCH) system aboard the International Space Station. The technology is designed to capture, relocate and remove space debris using a robotic system with tentacle-like arms and gecko-inspired adhesive pads. The ISS National Laboratory provided opportunities for over 170 capture trials, allowing engineers to refine the technology. REACCH could be used to steer satellites away from collisions, extend satellite life through servicing missions, and guide obsolete spacecraft towards controlled deorbiting.
The US Federal Communications Commission has updated its "Covered List" to include foreign-produced advanced robotic devices and power inverters, following determinations by a White House-convened interagency body that these products pose unacceptable risks to national security. The new additions are due to concerns about supply chain vulnerabilities and cybersecurity risks, which could be exploited by foreign governments or malign actors. Producers of affected devices can apply for Conditional Approval from the Departments of War and Homeland Security to receive FCC equipment authorizations, while existing models are unaffected.
A company's robotic hands, meant to be impressive but not necessarily "sexy," sparked debate among tech enthusiasts. Meanwhile, other AI-related advancements, such as a new technique that has helped improve the capabilities of language models and a promising approach to chipmaking, are gaining attention. These developments demonstrate progress in various AI areas, including natural language processing, artificial intelligence for science, and computing hardware.
Registration is now open for RoboBusiness 2026, taking place on October 20-21, 2026, at the Santa Clara Convention Center. The event will feature expert insights and innovative applications in robotics and AI across various sectors. Confirmed speakers include Rodney Brooks, Ken Goldberg, Jonathan Hurst, Bhavna Chandrashekhar, Morgan Pope, and Dermott O’Driscoll, among others. Over 100 exhibitors will showcase the latest enabling technologies, products, and services on the show floor.
Agency Tool Company's new startup seeks to build fast, reliable infrastructure for robotics developers by launching ATC Deploy, a software update tool that ships software to robots wherever they are. The system works with existing deployment strategies and architectures, allowing customers to continue using their own operating systems, docker images, and file structures. ATC Deploy aims to address the industry's bottleneck on software updates, which can be affected by flaky connections and limited resource availability on devices.
Integrated actuators improve humanoid robot joint performance and system integration by providing a compact solution that combines multiple components into one module. These modules offer advantages such as reduced weight, simplified wiring requirements, and improved mechanical layout. The AK45-10 actuator, for example, supports humanoid robots with its 7Nm peak torque output through a 10:1 planetary gearbox.
A team of scientists in Europe has developed a robotic fingertip with synthetic skin that reflects different colors of light in response to mechanical deformation, allowing it to generate high-resolution maps of topology, strain, and contact pressure. The skin uses a Bragg reflector, which is deformed by an object to change the reflected wavelengths. This technology provides 100-micrometer resolution with no computational latency, making it a promising option for roboticists to experiment with. The sensor embeds tactile sensing directly into the material of the finger, translating the deformation into digital signals without using traditional taxels or pixels.
Zenoh is a lightweight communication layer designed for ROS 2 that aims to reduce latency, minimize bandwidth usage, and simplify communication across distributed systems. It provides advantages for robots running on resource-constrained hardware or requiring cross-network communication. The author plans to explore Zenoh's behavior in practical robotics scenarios, including comparing it with the default DDS-based setup, over the next few weeks.
Neura Robotics has partnered with RWTH Aachen University to establish a new Neura Gym in Germany, scheduled to open by the end of 2026. The facility will be part of a worldwide network of Physical AI training centers where robots can be trained using real-world environments and simulation to generate data for AI development. Ten Neura Gyms are currently under development, with five expected to be operational across Europe, the US, and China by the end of 2026, addressing the shortage of real-world training data in Physical AI.
Tacta Systems has unveiled a robotics platform, TactaBot, designed to bring human-like dexterity and touch to industrial automation. The platform combines a robotic hand, a wearable skill capture system, and an artificial intelligence engine that enables robots to perform complex manufacturing tasks requiring precision, adaptability, and tactile feedback. At the centre of the platform is the Tacta Hand, which features 15 independently actuated joints and pairs with tactile sensors to detect force and temperature. The integrated platform aims to automate tasks traditionally difficult for robots, including precision electronics assembly and assembling wire harnesses.
American Industrial Partners has completed the acquisition of Honeywell Technologies Warehouse and Workflow Solutions, including Intelligrated and Transnorm. The combined organization, led by CEO Alfred Rebello, operates as a single entity focused on providing world-class warehouse automation solutions for retail, e-commerce, manufacturing, distribution, and parcel operations. The businesses offer a comprehensive portfolio of equipment, software, and services to help customers optimize operations, modernize fulfillment networks, and prepare for future challenges. The combined organization brings over $1 billion in revenue and employs more than 3,700 people across North America, South America, Europe, and Asia.
ENVue Medical Inc. has developed ENvue Drive, a robotic feeding tube placement system that uses artificial intelligence, real-time navigation, and robotic-assisted procedural execution to automatically place feeding tubes into patients. The system is designed to reduce complications, such as misplaced tubes ending up in the lungs, which can lead to serious health issues. ENvue's system provides electromagnetic guidance during enteral feeding tube placement and includes a sensor at the end of the feeding tube that gives physicians visibility while placing the tube.
NoTip has launched a public beta of its robotic food delivery platform in Mason, Ohio. The company's dual-stage delivery system uses an electric road vehicle to transport orders before a robot completes the final doorstep delivery. Residents can download the NoTip app and participate in the beta programme, gathering customer feedback on the service. Customers pay a flat $6 delivery fee with no tipping or restaurant price mark-ups.
Sapphire Technology has launched the EDGE+ Apex SOM/Carrier Robotics Platform, a production-ready hardware platform powered by AMD Ryzen AI Embedded X100 Series processors for physical AI applications in autonomous robotics. The platform delivers perception, planning and control performance with up to 128GB of LPDDR5x memory and a purpose-built robotics carrier board. It is designed for continuous industrial operation and supports an open software ecosystem for developers.
Researchers developed the SWiFT framework, a Swimming With Flow Toolbox, to enable efficient exploration of egocentric station-holding policies for robotic fish in unknown turbulent background flows via reinforcement learning. The policy achieves substantial improvements over state-of-the-art methods, particularly in root-mean-square error (RMSE) of distance. Egocentric feedback alone enables station-holding without explicit flow sensing, mirroring the biological phenomenon of rheotaxis.
A double-track vehicle model incorporating four-wheel differential actuation is established, which enables a drift-equilibrium calculation method and a closed-loop drift controller to be developed. The proposed approach achieves steady circular drifting with a sideslip angle of approximately 20° and performs figure-eight drift tracking through differential wheel torques in a steering-free four-wheel independently driven vehicle architecture. This study demonstrates the feasibility of drift control using only differential wheel torques for near-limit control in such vehicles, providing a new perspective on autonomous driving near handling limits. The approach is validated through simulations and experiments on a 1:10-scale vehicle.
Researchers have introduced a novel approach to differentiate the stationarity residual at the tolerance-converged solution in differentiable simulation, utilizing the Implicit Function Theorem (IFT). This method avoids expensive and step-size sensitive finite differences and intricate KKT sensitivity derivations. It requires nearly constant compiled temporary memory, growing by less than 4% with solver effort, compared to unrolled automatic differentiation.
FIRMGrasp, a family of friction-volatility-aware grasp quality metrics, predicts failures in force closure across the range of friction values. Unlike classical metrics, it evaluates the force-closure margin at the Conditional Value-at-Risk-discounted mean of the adverse friction tail. FIRMGrasp establishes its monotonicity and differentiability in grasp parameters and provides a probabilistic closure certificate guaranteeing force closure with probability at least β.
Előd Páll and Oliver Brock introduce Environmental Constraint Exploitation (ECE), a motion planning approach that simplifies planning by deliberately contacting the environment, reducing dimensionality and computational complexity. ECE integrates into motion planning algorithms to bias exploration to task-relevant regions and leverage contact for uncertainty reduction. The authors evaluate ECE benefits using RRT-based planners and demonstrate practical benefits in a real-world application.
A hybrid approach combining Artificial Potential Fields (APF) with a Spatio-Temporal Transformer achieved superior balanced performance in real-time Autonomous Underwater Vehicle (AUV) path planning, outperforming standalone learning models and classical methods. This framework demonstrated the shortest average path length, lowest collision rate, and efficient computation time while ensuring reactive obstacle avoidance and global path optimality. It offers a robust solution for real-time AUV navigation, harmonizing obstacle avoidance with path optimization in resource-constrained underwater systems.
A cable-driven continuum robot arm, which is underactuated and prone to residual vibration during rest-to-rest maneuvers, utilizes a time-delay filter as an input shaper to eliminate vibratory mode excitation. The non-robust and robust filters demonstrate improved response compared to velocity-driven pulse inputs. Experimental results using the continuum robot validate the application of the input shaper, showing reduced overshoot and settling time.
Researchers at Saayuj Deshpande, Gia Dcosta and Samhitha Vedire investigate Star-World workspace reshaping for three-dimensional reactive control of a Franka Emika Panda manipulator. At each update, intersecting inflated obstacles are clustered and replaced by star-shaped proxies before a dynamical-system-based end-effector controller is evaluated. The method resolves the canonical overlapping-wall case and requires 0.68--8.70\,ms per workspace update for scenes containing one to seven obstacles, but increases path length in some cases and produces near-equilibria.
A lightweight sensor-driven Lévy walk (SDLW) controller enables efficient autonomous exploration with palm-sized nano-UAVs. The method uses a sensor-reactive heading policy to diversify exploration without inter-robot communication. Simulation results show improved coverage and reduced collisions compared to a uniform-heading Lévy walk baseline, making it suitable for scalable multi-UVAE exploration on minimal-sensing platforms.
A method leverages semantic maps from OpenStreetMap to enable robots to localize in previously untraversed environments against prior data. The approach involves extracting relevant landmarks from panoramas and identifying feasible correspondences between these landmarks and prior overhead landmarks using Variational Likelihood Machines (VLMs). A lightweight matcher is then distilled from the VLM to compute correspondences for all entities in a map, which is fused with a Bayes filter to create a time series of pose estimates. The method generalizes across location, lighting, weather, and other challenges, and code and datasets are available.