Hexagon AEON Humanoid Robot Demos Nvidia AI Powered 34 DoF Tech (GENERALIST + Q5)

Hexagon’s robotics division has introduced Aeon, an autonomous robot with advanced sensors and AI, designed to help with labor shortages and improve safety in various industries. It can perform a wide range of tasks, make decisions quickly, and work efficiently alongside humans. Meanwhile, Generalist, a robotics company, is developing robots with fine motor skills and adaptability, capable of handling diverse tasks. Lastly, Robot Era has revealed Star Q5, a humanoid robot with high dexterity and compactness, designed for both industrial and everyday environments.

Hexagon’s robotics division has unveiled Aeon, a new autonomous robot designed to address labor shortages and enhance safety in a variety of demanding industries. At the core of Aeon’s capabilities is Hexagon’s advanced sensor suite, combined with a powerful AI-driven mission control system and spatial intelligence. This trio allows Aeon to take on a broad set of real-world tasks, from delicate manipulation and asset inspection, to developing digital twins and supporting human operators on the factory floor. Aeon’s mobility is achieved through a wheeled base that allows for efficient forward, backward and pivoting motion. Its manipulation system is modelled on bionic hands, offering high dexterity for complex tasks.

Internally, Aeon houses three specialized computers, a general computing engine and two dedicated AI processors for edge-based autonomy and advanced mission reasoning. The emphasis on edge AI means Aeon can process information and make decisions locally, without relying on constant cloud connectivity. This reduces latency and allows the robot to react swiftly to changes, whether navigating a busy warehouse or inspecting critical infrastructure. Plus, Hexagon’s focus on safety and support means Aeon can work alongside human colleagues reliably and efficiently. On top of this, Aeon’s agility stands out. It leverages Hexagon’s proprietary precision measurement technologies, enabling swift navigation and high-accuracy work.

Plus, Aeon’s spatial intelligence means it fuses data from a network of 22 sensors, including 12 cameras for peripheral, front-facing, rear and augmented reality heat detection, providing an acute awareness of its environment. Furthermore, Aeon’s design emphasizes versatility. The robot can pick specific objects, scan components for inspection, and create digital replicas of environments, all through an end-to-end training approach. This makes it a fit for tasks ranging from reality capture to teleoperation. Another noteworthy feature is its autonomy and power. Aeon runs on two self-swapping batteries, eliminating downtime for recharging and supporting continuous operation.

Aeon’s advanced capabilities are built in partnership with leading technology companies. The robot’s computational backbone includes NVIDIA’s accelerated computing platforms, Omniverse and Jetson modules. This enables advanced AI processing both at the edge and in the cloud. Built to operate across sectors such as automotive, aerospace, manufacturing, warehousing and logistics, Aeon blends state-of-the-art hardware and cutting-edge artificial intelligence into a single adaptable platform. Meanwhile, California-based robotics company Generalist has taken another significant stride toward realizing its mission of deploying general-purpose robots capable of handling a wide array of real-world tasks. The company recently released a series of capability demonstrations, highlighting its progress in developing dexterous end-to-end AI models for autonomous manipulation across diverse environments and physical interactions.

At the core of Generalist’s approach lies the use of end-to-end deep neural networks, which map raw sensor data, such as camera pixels and force readings, directly to robot actions at a rapid 100 Hz control rate. This tightly integrated hardware and software stack allows for smooth, reactive manipulation, enabling robots to perform complex tasks that demand both fine motor skills and robust adaptability. In these latest demonstrations, the robots complete tasks that not only test the limits of dexterity, but also evaluate their generalization across various robotic embodiments and environments. For instance, one task requires the robot to pick and sort small fasteners from a cluttered pile, orienting and placing them into designated compartments.

Here, speed is bounded primarily by the hardware’s torque limits, but the AI model’s ability to generalize across clutter and object variability is especially noteworthy. Another scenario showcases the robot folding a cardboard box, packing a flexible bicycle chain lock inside, and then closing the box. This sequence stresses the robot’s capacity for handling articulated and deformable objects over long action horizons. The system must precisely modulate gripping force, strong enough to manipulate the materials, yet gentle enough to avoid crushing the box, and deal with the delicate alignment of the box flaps, each requiring millimeter-level precision.

Plus, generalists’ robots demonstrate by manual coordination, tool use, and dynamic re-grasping in tasks such as returning screws to a glass jar. The robot adapts tactics on the fly, sometimes scraping screws off a magnetic holder, other times improvising by bending a paper plate into a funnel to pour them in. This requires precise timing between both hands, as well as the ability to handle shiny, small objects and transparent containers, a combination that typically challenges machine vision systems. On top of this, the robots exhibit high-velocity maneuvers in the Lego challenge, breaking apart, sorting, and even throwing Lego bricks into color-coded bins.

This task tests quick re-grasping, forceful interactions, and rapid adaptation, as the robots must deal with a variety of brick formations and bin placements, relying on visual input for real-time decision-making. The model generalizes to entirely new layouts and robotic arms, underscoring the system’s flexibility. A distinctive aspect of generalists’ research is cross-embodiment transferability. Their AI models are trained in ways that allow seamless deployment across different robotic platforms, such as the 7-degree-of-freedom Flexiv Reson 4 and the 6DOF UR5 arm. In some cases, the system performs tasks in new environments and on new hardware without having seen task-specific data during training, a promising indicator of the technology’s robustness and generalization.

Generalists’ robots achieve this level of dexterity and autonomy due to the marriage of high-frequency control, continuous sensor feedback, and unified neural network policies. This combination enables real-time adjustments, whether it’s wiggling objects for better grip, throwing, or delicately closing a container. Furthermore, the system’s ability to adapt to external disturbances and dynamically modulate force makes it suitable for unpredictable and unstructured environments. Finally, Robot Era has officially unveiled its latest innovation in humanoid robotics, the Star Q5. Designed with a focus on dexterity, compactness, and intelligent interaction, the Q5 aims to set a new standard for robots operating in both industrial and everyday environments.

The Star Q5 features wide-range movement and dexterous manipulation capabilities, thanks in part to its 11-degrees-of-freedom and a 7-axis high-precision anthropomorphic arm. Each hand, which is slim and closely matches the size of a human hand, can support a load of up to 10 kilograms and achieves fast response rates of 10 clicks per second. On top of this, the Q5 leverages micro-force control joints for precise nuanced handling, making it well-suited for tasks requiring both strength and delicate manipulation. Compactness is a major highlight of the Q5. It’s built to maneuver easily in tight spaces and on narrow pathways.

This is further supported by an intelligent navigation system that combines 3D LiDAR and binocular camera vision, enabling autonomous movement and situational awareness in complex environments. Interaction is another key component, with the Q5 providing a hyperanthropomorphic form factor and supporting anthropomorphic dialogue. Its unique waist design and lifelike appearance make human-robot interactions more natural. Furthermore, the Q5 allows for precise, synchronized whole-body teleoperation, with compatibility for data gloves, VR, and other remote control technologies, making it suitable for applications ranging from remote assistance to hazardous environment work. The robot comes integrated with a StarPower smartphone and operates on RobotEra’s ERA AI platform, an end-to-end embodied AI system.

Overall, the StarQ5 positions itself as a flexible, intelligent and robust robot, geared for safe, effective operation in diverse environments, plus, with its precise control and advanced teleoperation features, it promises to be a versatile addition to the growing humanoid robot market. [tr:trw].