PNDbotics pushes Adam-U as a stationary humanoid built for manipulation data, teleoperation training

Reported specifications put Adam-U at an adjustable height of 135 to 177 cm, with a 60 cm width and 40 cm depth, making it compact enough for lab workstations while still resembling a human upper-body workspace. Its 61 kg weight and stationary platform format support stability during arm and hand work, while the listed sensor package includes binocular vision, IMUs, tactile hand sensing, force sensors in the actuators, and motion-capture integration. PNDbotics also says the system is built around a ROS-based OS with C++ and Python access, and the company’s CES material points to an emphasis on simulation and real-world transfer rather than locomotion.

Adam-U reflects a broader shift in humanoids from pre-programmed motion libraries toward systems that learn from human demonstration and simulation. That matters because it reframes success: the near-term value of a humanoid may come from how efficiently it can collect, replay, and refine behavior, not from whether it can handle every task on its own. For buyers in research and industrial training, that is a more measurable product category than a humanoid that promises general intelligence before the software is ready.

Humanoid robots are increasingly being judged less by how far they can walk and more by how well they can mirror human motion, hold tools, and generate useful training data. PNDbotics has been showcasing Adam-U as a stationary humanoid research platform for data acquisition and operation training, highlighting its 31 degrees of freedom, dexterous hands, and motion-capture-oriented setup at recent CES appearances.

Adam-U’s pitch is narrower than a general-purpose humanoid’s, and that is the point. By focusing on full-body motion imitation, balance correction through a waist brake system, and hand-level tactile sensing, the platform is aimed at researchers who need repeatable manipulation rather than open-ended autonomy. Its ROS-based software stack, C++ and Python APIs, and Isaac Gym compatibility also matter because they lower the friction between simulation, teleoperation, and machine learning workflows.
Adam-U is designed less as a walking robot and more as a data engine for humanoid control.

For Adam-U, the system flow is straightforward: human motion input goes into AI model processing, which then drives joint actuation and balance correction. In practice, that means the robot is meant to capture what a person does, translate it into coordinated movement, and use onboard sensing to keep the torso and arms stable while manipulating objects. PNDbotics has also tied the platform to Sim2Real workflows, suggesting a loop where actions are learned in simulation and then transferred to the physical robot.

The most concrete use case is robotics research inside labs and R&D centers, where Adam-U can be used for dexterous manipulation tasks such as tool handling, object sorting, and operation training. That setting fits the platform’s stationary design: a fixed base reduces locomotion complexity and lets teams concentrate on manipulation quality, teleoperation fidelity, and data collection for imitation or reinforcement learning. The result is a robot optimized for controlled human-centric spaces rather than open environments.

Adam-U

Leju Robotics positions KUAVO 5 as a teleoperation-first humanoid for factories and service spaces

KUAVO 5 is reported at 173 cm tall and 63.5 kg, a size and mass range that make it look and move more like a person than a mobile machine. Its listed movement profile includes 0.4 m/s walking and 5 km/h running, which supports short-range movement in shared indoor spaces, while the sensor suite combines RGB-D/depth vision, LiDAR, a 6-microphone 360° array, an IMU, joint temperature sensors, and stereo speakers. The platform is also reported to run on a KaihongOS/OpenHarmony-based stack, tying embodied AI and real-time control into the same software foundation.

KUAVO 5 also reflects a broader shift in humanoid robotics: near-term buyers appear to want robots that can be deployed with human supervision before autonomy matures. That is a meaningful market signal because it favors platforms that can handle imitation, remote guidance, and shared-space safety over robots built mainly around pre-programmed motion libraries. In that sense, Leju’s robot is part of a practical turn in the category, where the first commercial value may come from extending human labor rather than replacing it.

Humanoid robots are being pushed toward tasks that require balance, human-scale movement, and object handling in spaces built for people. Leju Robotics’ KUAVO 5 was presented as the latest generation of its full-size humanoid platform, with demonstrations and product reporting emphasizing bipedal motion, modular hardware, and remote-operation use cases.

What separates KUAVO 5 from many humanoid demos is the way its reported design centers on teleoperation, full-body coordination, and manipulation rather than scripted motions. Its modular body and reported 5G-A remote teleoperation support point to a robot intended for human-in-the-loop work in factories, service areas, and inspection environments. The platform also appears built around natural interaction, using a sensor stack that includes depth vision, LiDAR, microphones, and speakers to support awareness and communication. KUAVO 5’s main story is not autonomy alone, but how closely it can map human intent onto a stable humanoid body.

The system flow is straightforward: human motion input goes into AI model processing, and the robot converts that into joint actuation with balance correction. In practical terms, that means an operator can guide the robot while onboard sensing and control keep the body upright, aware of obstacles, and responsive to uneven movement. For a humanoid category, that pipeline matters because the hardest problem is not just moving forward, but moving like a person while staying safe around people.

A realistic deployment scenario is assisted material handling on a factory floor, where KUAVO 5 can move packages or tools between stations while a remote operator supervises the action. That use case fits the robot’s reported bipedal mobility, manipulation-oriented upper body, and safety features such as collision detection, force limiting, and fall recovery. It also matches the category’s current operating model: humanoids are often most useful when they extend human reach in spaces designed around human workflows, not when they replace every step of the job.

KUAVO 5

Techman Robot’s TM Xplore I debuts as a wheeled humanoid aimed at factory work and teleoperation

The publicly reported 167 cm height gives TM Xplore I a human-scale profile that should help it operate around workbenches, carts, and factory fixtures designed for people. Its reported top speed of 0.83 m/s, or about 3.0 km/h, points to careful indoor movement rather than fast transit, which is consistent with collaborative use in shared spaces. Techman has not publicly disclosed its weight, and that omission leaves durability, transportability, and balance margins as open questions.

TM Xplore I also points to a broader industry shift toward teleoperation-first humanoids and away from the idea that robots must be fully autonomous before they are useful. By pairing a humanoid upper body with a wheeled base, Techman is effectively betting that controlled movement, remote supervision, and task-specific learning will reach customers sooner than fully general-purpose walking robots. That is a meaningful signal for manufacturers: the next commercial humanoids may be judged less by cinematic demos and more by how reliably they fit into existing production lines.

Humanoid robots are being pushed into the gap between repetitive industrial work and tasks that still depend on human judgment, especially in shared spaces where balance, dexterity, and safe interaction all matter. Techman Robot unveiled its TM Xplore I humanoid at NVIDIA GTC 2026, presenting a wheeled platform with an upper body designed for manipulation, visual inspection, and autonomous movement in factory settings.

What distinguishes TM Xplore I is not raw speed, but the way Techman is pairing a humanoid torso with a wheeled base and edge AI to target stable movement, human-level interaction, and practical task handling in manufacturing environments. The company’s emphasis on multimodal sensing, autonomous navigation, and quick adaptation to precision work suggests a robot built for assisted labor rather than theatrical demos. That positioning matters because it aligns humanoid development with real factory workflows, where reliability and controlled movement tend to matter more than acrobatics.
TM Xplore I is less about mimicry and more about usable human-robot coordination.

The reported system flow is straightforward: human motion input and environmental data are fed into AI models, the software interprets the task, and the robot then adjusts joint actuation and balance in real time. Techman says the platform uses its “See, Think, Act” stack alongside NVIDIA Jetson-based edge computing, while source coverage also points to vision-language-action modeling for sensor fusion and navigation. In practice, that means the robot is being framed as a machine that can perceive a scene, decide how to move, and execute manipulation without relying on fixed motion scripts.

The clearest near-term deployment scenario is supervised factory handling, where TM Xplore I can move between stations, carry parts or tools, and assist workers with inspection or assembly support. That use case fits the robot’s wheeled design and its reported emphasis on stability, because industrial floors reward predictable travel, short-range manipulation, and quick adjustments when a person steps into the robot’s path. The strongest commercial logic is in human-centric production zones where teleoperation or assisted autonomy can reduce the need for workers to perform physically awkward or visually intensive tasks.

TM Xplore 1

Zoomlion’s Z01 humanoid lands at KOMATEK with a Tai Chi demo and an industrial pitch

Zoomlion has not publicly disclosed Z01’s dimensions, weight, speed, or battery life, so those performance characteristics remain unknown. What is reported is its sensor-and-control profile: RGB cameras, depth cameras, IMU, gyroscope, force and torque sensing, joint encoders, and collision/contact sensing, with possible LiDAR support for navigation. The navigation stack is described as visual SLAM and indoor SLAM, while the software layer is presented as a proprietary embodied-AI system with likely ROS or ROS2 integration and task-level APIs.

Zoomlion’s presentation also points to a broader shift in humanoids toward teleoperation-first deployment, where robots are judged by how reliably humans can direct them in real time. That matters because the near-term market appears to reward systems that can work in controlled indoor environments with human oversight, rather than platforms that promise broad autonomy before the hardware and software stack is mature. Z01 fits that pattern: a humanoid framed as a practical interface between embodied AI and human workflows, not a fully independent worker.

Zoomlion Heavy Industry Science & Technology unveiled its humanoid robot Z01 at KOMATEK 2026 in Istanbul, where the robot performed a Tai Chi routine and drew attention as a bipedal platform for industrial collaboration and guided interaction. The demonstration framed Z01 as part of Zoomlion’s broader embodied-AI push, with the company positioning the robot for human-centric settings where motion control, balance, and close-range interaction matter.

Z01 matters because it is being introduced as a humanoid built around full-body coordination rather than isolated task automation, which is the central challenge for legged robots in real spaces. The demo emphasized real-time balance correction, natural movement, and interaction readability, all of which are more relevant to teleoperation and assisted service than to pre-programmed stage choreography. Zoomlion is also signaling a product strategy that ties humanoids to embodied-AI workflows instead of treating them as standalone showpieces. Z01’s value is less about novelty than about making human-directed robot motion look operationally usable.

For a humanoid like Z01, the core flow is human motion input, AI model processing, then joint actuation with balance correction, which is the sequence implied by its live movement and interaction demos. In practical terms, the robot appears designed to take visual and motion cues, interpret them through an embodied-AI control stack, and translate them into coordinated leg, arm, and torso actions while keeping its posture stable. That kind of pipeline is what allows a humanoid to imitate gestures, respond to operators, and stay upright in close human environments.

The most concrete deployment case is intelligent tour guidance in indoor, human-dense spaces, where a humanoid can combine movement, gesture, and proximity-aware interaction without needing factory-scale autonomy. Z01 was explicitly described as suitable for intelligent guidance and educational applications, which fits scenarios such as exhibition halls, corporate demo floors, and training venues where the robot’s presence is part of the experience. In that setting, its main job is not heavy lifting or long-range autonomy but readable, socially legible motion under human supervision.

Robot News & Industry Updates

One Robot
Infinite Possibilities

One Robot
Infinite Possibilities

TP-01 by Raydiculous

PNDbotics pushes Adam-U as a stationary humanoid built for manipulation data, teleoperation training

Latest Articles

Leju Robotics positions KUAVO 5 as a teleoperation-first humanoid for factories and service spaces

Techman Robot’s TM Xplore I debuts as a wheeled humanoid aimed at factory work and teleoperation

Zoomlion’s Z01 humanoid lands at KOMATEK with a Tai Chi demo and an industrial pitch

More Robot Stories

Robot News & Industry Updates

One RobotInfinite Possibilities

One RobotInfinite Possibilities

TP-01 by Raydiculous

PNDbotics pushes Adam-U as a stationary humanoid built for manipulation data, teleoperation training

Latest Articles

Leju Robotics positions KUAVO 5 as a teleoperation-first humanoid for factories and service spaces

Techman Robot’s TM Xplore I debuts as a wheeled humanoid aimed at factory work and teleoperation

Zoomlion’s Z01 humanoid lands at KOMATEK with a Tai Chi demo and an industrial pitch

More Robot Stories

One Robot
Infinite Possibilities

One Robot
Infinite Possibilities