Musk Praises Unitree G1; LimX Dynamics Unveils Multi-Form Robot; Daxiao Robotics Showcases Three Major Achievements | Electric Wave Weekly Report

1. Musk praises Unitree G1; appears at Wang Leehom concert

On December 18, the Unitree G1 robot made an appearance as a backup dancer at Wang Leehom's Chengdu concert, performing together to "Full Power." The performance video went viral internationally, even catching the attention of Tesla CEO Elon Musk, who commented: "Impressive."

In the video, six G1 units wore silver sequined tops and black leather pants on stage. The G1 robots replicated every move of the dancers with smooth execution throughout the performance, culminating in a collective backflip that electrified the audience.

This year, robots have frequently entered the public eye, particularly through applications in the entertainment industry. Various robotic personas are becoming more vivid, delivering emotional value to audiences in entertainment settings.

Of note, the G1 robots featured at Wang Leehom's concert were provided by Unitree Robotics' official team. Previously, Unitree Robotics founder Wang Xingxing had personally demonstrated the robot's dancing capabilities to Wang Leehom.

2. LimX Dynamics officially releases multi-form robot TRON 2

On December 18, LimX Dynamics launched the new multi-form Embodied AI robot TRON 2, with a starting price of 49,800 yuan. The robot features a fully modular architecture and can be configured into three forms: dual-arm, bipedal, and wheel-legged.

In its dual-arm configuration, TRON 2 has a payload capacity of 10 kg for its arms and a maximum load capacity of 60 kg for the body, meeting the pick-and-place and material handling needs in industrial scenarios.

Additionally, in this form, TRON 2 is compatible with the large model Pi 0.5 and ACT. Based on home operation scenarios, it covers 10 typical operational tasks including meal preparation, tabletop cleaning, clothing care, item storage, and package retrieval, providing datasets, model training, and algorithm deployment examples.

In both bipedal and wheel-legged modes, TRON 2's all-terrain mobility is enhanced, offering up to 4 hours of battery life and a maximum load capacity of 30 kg.

As seen in the video, the bipedal version of TRON 2 can walk up stairs, while the wheel-legged version boasts rapid movement capabilities, making it suitable for smooth travel on flat surfaces.

3. Dash Robotics Showcases Three Major Achievements

On the afternoon of December 18, Wang Xiaogang, co-founder and executive director of SenseTime, and chairman of Dash Robotics, publicly unveiled the latest progress of Dash Robotics for the first time. He also announced the ACE embodied R&D paradigm, the Kaiwu World Model 3.0, and the hardware product—the Embodied AI Super Brain Module A1.

One of the core components of the ACE paradigm is environment-based data acquisition technology. This data acquisition method integrates multi-modal data such as first-person and third-person videos, force-tactile information, motion trajectories, and voice to build a physics-based 3D asset library. It covers various interaction information required for training embodied models. Currently, this solution has achieved multi-view data acquisition, enabling tasks that are difficult to realize under teleoperation paradigms in scenarios like homes and retail stores, such as fresh produce grasping and precision placement.

The Kaiwu World Model 3.0 features an integrated architecture of 'multi-modal understanding—generation—prediction.' By inputting multi-dimensional information including vision, 3D trajectories, touch, and friction, it deeply understands the physical laws of the real world and the underlying logic of human behavior. It is reported that the environment-based data acquisition in the ACE embodied R&D paradigm can collect tens of millions of hours of data per year. The Kaiwu World Model 3.0 continuously amplifies the value of real-world data, achieving the effect of scaling up to hundreds of millions of hours of data.

Additionally, the Embodied Super Brain Module A1 can be mounted on quadruped robots to help solve behavioral autonomy issues, allowing them to adapt to complex and unfamiliar environments to achieve autonomous movement.

In specific applications, quadruped robots equipped with A1 can identify uncivilized behaviors in parks and detect dangerous situations, transmitting information back through cloud-based intelligent interactions.

4. Qianxun Intelligent Robot Deployed at CATL Battery Production Line

On December 18, Qianxun Intelligent announced that its humanoid robot, "Xiao Mo," has officially begun operations on the PACK production line for CATL's new energy power batteries, successfully completing complex tasks such as precise connector insertion.

Xiao Mo is equipped with a VLA (Vision-Language-Action) model, enabling environmental perception and task generalization. It performs reliably in the final functional testing procedures for battery packs before EOL (End-of-Line) and DCR (Dynamic Capacity Rating) discharge. In actual operation, its connector insertion success rate remains stable at over 99%.

During operation, Xiao Mo can autonomously detect wire harness connection status, immediately reporting any anomalies. It can also proactively switch to an inspection mode during work intervals. For continuous production tasks involving multiple battery models, its daily workload has increased threefold.

5. Tanshi Zhizhang Releases Full-Stack Technology

On December 19, Embodied AI company Tanshi Zhizhang released its full-stack technology, which includes human-centric embodied data and model paradigms, the TARS AWE 2.0 model, and a hardware system designed specifically for AI.

During the release segment, itStone showcased the world's first robot capable of completing hand embroidery. Through the process of threading the needle and stitching, it ultimately embroidered the company's logo.

In an industry facing challenges in achieving zero-distance collaboration between software and hardware, itStone has provided a systematic solution integrating DATA, AI, and PHYSICS into a triad.

First is the human-centric embodied data and model paradigm. Under this paradigm, through hardware-level innovations, itStone has built a lightweight, fully modal, wearable embodied data acquisition system called SenseHub, which can continuously record real, high-quality operational behaviors.

Regarding models, itStone has constructed TARS AWE 2.0 based on massive real-world data. Through spatial perception pre-training, it builds a cognitive model of the world for robots, whose generalization capabilities enable core skills to be efficiently transferred across different scenarios.

On the hardware body aspect, itStone displayed the A series (top left) and T series (top right) body robots. Adhering to the core design principle of minimizing the 'digital-to-physical gap,' these systems are equipped with a full suite of self-developed core components and super sensor combinations, enabling algorithmic capabilities from the digital world to be accurately and stably mapped to real-world physical operation scenarios.