And this is crucial when combined with the robot’s new vision-language model, which now offers improved interpretation of visual data received by its cameras, resulting in more informed decisions and the ability to carry out long-horizon tasks. Plus, by combining its VLM with six onboard RGB cameras, it equates to an intelligence system that’s much more human-like in its awareness of surroundings. This is because two cameras are located in its animated face, with its additional cameras distributed around its body for an AI-driven vision and execution system that enables completely autonomous task automation. And that’s only the beginning, because Figur 2 also boasts a three-fold increase in computational power compared to its predecessor, allowing it to perform complex AI tasks with increased understanding and precision.

Furthermore, its exterior features sleek body panels with internal wiring to prevent entanglements and minimize risks when working with humans. It also integrates newly engineered joints that are more powerful than ever before, with shoulder joints boasting 50 Nm of torque and a 148-degree range, knee joints with 150 Nm and 135 degrees range, and hip joints with a similar 150 Nm but 195 degrees range of motion. On top of this, the robot’s hands have also been upgraded to perform a wide range of human-like tasks with 16 degrees of freedom and a lifting capacity of 20 kg to match.

But what’s even more important is Figur 2’s addition of a new 2.25 kWh custom battery pack incorporated into its torso, reportedly increasing energy delivery by over 50% to operate for five hours at a time in real-world work environments. But specific runtime figures around battery charge time have not been disclosed, and Figur 2 has already been deployed at BMW’s manufacturing facility in Spartanburg, South Carolina, following the training success of its predecessor. Importantly, though, the newer model has already begun AI data collection and use case training on the factory floor. As for the future, Figur’s goal is to create robots that will not only enhance productivity and efficiency, but also improve the quality of life for humans both inside and outside of the workplace.

And while no price tag is currently stated for the robot, it appears that this robot will likely cost somewhere between $40,000 and $90,000 upon release. Meanwhile, UBTEC Robotics from China has also just announced a new real-world humanoid initiative with automotive giant Geely Holding Group and Miracle Automation Engineering. Specifically, the goal is to integrate UBTEC’s robots into intelligent manufacturing for automobiles and components, marking a milestone in the adoption of advanced robotics in the automotive industry. In fact, Walker S. Light has already been deployed at a separate 5G-enabled smart factory, where for 21 days, Walker S.

Light worked alongside human employees while packing and securing cargo in the factory’s cargo transport unit. And during its three-week training period, the robot utilized cutting-edge technologies like semantic V-Slam navigation, end-to-end imitation learning, visual precision recognition, and full-body fine motion control. These advancements allowed Walker S. Light to autonomously navigate and transport payloads to various pallets and assembly lines, collaborating with an automated guided vehicle to fulfill its tasks. And under UBTEC’s new partnership agreement, Geely has agreed to provide application scenarios for the Walker S. robot, while Miracle Automation will also offer engineering expertise in data collection, process model development, and scene improvement design for the robot’s applications and training.

Altogether, UBTEC is rapidly expanding efforts to commercialize its humanoid robots in industrial smart manufacturing across China, with the company having already collaborated with car manufacturers like Nio, Dongfeng, and Volkswagen to integrate Walker S. into their operations. And while there isn’t yet any pricing information around the Walker S. Light, the company says they’re planning on a wider rollout in 2025. And finally, Menti has just demonstrated its updated humanoid robot as it follows and assists humans in real-world scenarios in an exciting breakthrough showcasing the company’s latest AI advances. To start, the team behind the Menti is utilizing model distillation to run large foundation models efficiently on GPUs, ensuring high frame rates without sacrificing accuracy.

Additionally, Menti’s capabilities include converting 2D objects into 3D models using a combination of 3D point cloud and segmentation models like Nano-Sem. Furthermore, the robot employs neural radiance fields augmented with visual embeddings to enhance its understanding of the environment, allowing the robot to navigate complex spaces and perform tasks with remarkable precision. The robot uses an advanced localization system that fuses nerves with visual odometry, creating a topographic map for navigation. This map guides the robot in avoiding obstacles and executing tasks like object retrieval. By running algorithms like random sample consensus, Menti matches superpoints in nerf models with real-world images, ensuring accurate positioning.

On top of this, Menti’s 3D semantic model is enriched with information beyond RGB data, allowing the robot to perform open vocabulary searches, identifying objects like ovens or refrigerators without pre-defined lists. This system utilizes semantic features for enhanced querying to dramatically improve performance over traditional baselines. And what’s more is that Menti is also designed to integrate multiple components for diverse tasks, such as onboarding in new environments. The robot follows humans, learning from interactions, storing information for future use, and even employing AI agents to handle loosely defined tasks so it can adapt to events or improve object manipulation for actions like opening doors or calling elevators.

To ensure accuracy, the robot tracks and interacts with humans using advanced algorithms for person detection and re-identification. Menti’s pointing interface interprets gestures and verbal cues, distinguishing between locations and objects with the help of large language models. And because safety is paramount for Menti, the robot’s policies ensure both stability and adaptability, even on challenging terrains, employing models like S4 to predict and mitigate falls, or even transition into soft landings when necessary. Overall, the Menti’s humanoid robot exemplifies a bright future for commercial humanoid robotics by combining AI innovation with practical applications. In this way, Menti Robot is setting new standards in AI-driven robotics, and as it continues to evolve, humanoids like these are poised to transform the labor market and human life as we know it, and it’s happening much faster than the tech industry had previously expected.

And as we wait to see which humanoid will rise and rule, one thing is certain. The human labor market is about to face a very serious change of pace, and those who embrace AI first will likely be the participants who come out on top of this technological revolution. [tr:trw].