China's First Open-Source Humanoid Robot Has Arrived
China's first open-source humanoid robot

The significance of open source for the robotics industry is self-evident.
Previously, the full open-sourcing of MIT's Min Cheetah quadruped robot in the United States greatly boosted related technological innovation and industrial collaboration, substantially lowering the threshold for robot development. Quadruped robots flourished worldwide, with more and more R&D teams globally participating in the development and industrialization of related products and validating their ideas. This has undoubtedly accelerated global resource sharing and industrial coordination, contributing more possibilities for the advent of the robot era.
As the hottest robotics track today, humanoid robots are among the most anticipated products. However, the integrated development of humanoid robot hardware and software is extremely difficult. Building a humanoid robot that operates stably and efficiently requires mastery of technologies across multiple fields, including mechanical design, electronic engineering, and computer science, as well as in-depth research and development of perception and control algorithms—and it is costly. This has led to a general scarcity of researchers and developers, and has also limited subsequent research on artificial intelligence and machine learning based on humanoid robot platforms.

So, can an open-source humanoid robot in both hardware and software press the accelerator on humanoid robot research?
Overview of the Open-Source Product
Recently,Robotics Lecture learned that two Chinese companies, ENCOS Intelligence and BridgeDP, are planning to open-source a bipedal robot, including the underlying control algorithms and hardware structure. Robotics Lecture obtained exclusive product specifications for this robot.

In terms of appearance, this open-source humanoid robot product EC-Hunter80-V01 features a distinctive leg design,with no upper body or arms and removes currently unnecessary cosmetic components,retaining only the most basic structural parts so that the robot has no redundant parts and achieves maximum lightweighting. This is somewhat similar to Cassie, the humanoid robot prototype from Agility Robotics abroad, which features classic "ostrich legs" on its lower body,both embodying a minimalist mechanical engineering style. Cassie has since grown an upper body and arms to be upgraded into the Digit robot, with plans to produce 10,000 units per year.
In terms of hardware specifications,the EC-Hunter80-V01 stands 785.97 mm tall, with a front-to-back width of 178 mm and a side-to-side width of 330.4 mm. It has 10 degrees of freedom overall and a body weight of approximately 12 kg. The robot carries a 5,300 mAh lithium battery. Although the battery pack does not come with a unified battery management system, it can display power levels and issue low-battery alarms. The robot is equipped with two power line interfaces: 5V 10A and 19V 12.5A. The openness and compatibility of the control system and electric drive components are excellent.

For the control module,the EC-Hunter80-V01 uses the NUC wski7 control platform and an EtherCAT control interface. The EtherCAT interface employs a high-speed EtherCAT-to-CAN module independently developed by ENCOS Intelligence. One module driving 6 motors can achieve a communication frame rate of up to 1 kHz, while one module driving 2 motors can achieve up to 2 kHz.
Using an EtherCAT-to-CAN module conveniently solves the connection between Linux systems and joint motors. CAN communication offers high reliability, strong anti-interference capability, low cost, and small interface size. EtherCAT, on the other hand, uses RJ45 ports, which are bulky but can transmit large amounts of data at once and are easy to connect to systems such as Linux. Therefore, using EtherCAT-to-CAN as the connection between joint modules and the control platform on legged robots is a good choice. We also noticed that the robot's IMU is also an EtherCAT bus module independently developed by ENCOS,sharing the same bus as the motor control. This not only simplifies the overall hardware architecture but also ensures real-time communication.
The reason for having only the lower-body legs and feet is that ENCOS Intelligence believes that one of the main technical obstacles currently hindering the large-scale application of bipedal robots is lower-limb balance and locomotion, which is also a key technical challenge being tackled by major robot manufacturers worldwide. Therefore, on the hardware platform, the EC-Hunter80-V01 focuses on retaining the lower-body bipedal section,while also supporting free upgrades by users.

Algorithms are the core of humanoid robots, including motion control, perception, and decision-making. ENCOS Intelligence adopts a model-based control method and uses an MPC+WBC algorithm framework, integrating modules for localization and perception, gait planning, ground contact computation, and others to perform torque hybrid control on the legged robot.
ENCOS Intelligence revealed to Robotics Lecture that they plan to open-source the core control code of the EC-Hunter80-V01 robot, including low-level EtherCAT communication, bipedal motion control algorithms, gait planning, and debugging components. At the same time,the robot's complete mechanical structure and the hardware module solutions used will also be open-sourced.
This means that based on this open-source robot, users can modify and improve the robot's structure in hardware—for example, increasing the robot's degrees of freedom, adding toe joints on the feet, adding cosmetic components, and so on. Since the robot's payload reaches 15 kg, it can accommodate the addition of dual arms or even four arms, as well as vision systems.

At the software level, users can likewise improve the robot's control algorithm code on the basis of the open-sourced code—for example, by adding reinforcement learning algorithms to upgrade the robot's stability control. Overall,there is enormous room for modification and expansion.
In an interview, ENCOS Intelligence said it planned to open-source the robot's algorithms independently by February 8, 2024, ahead of Lunar New Year's Eve. The code would be hosted on GitHub. The repository is available at (https://github.com/bridgedp), while the hardware platform will be sold separately at an initial price of approximately 180,000 yuan.
The Foundation of Open Source
Achieving such a high degree of openness is made possible by the in-house development and production of the underlying motion control components.
In a previous Robotics Lecture report, "New Opportunities in the Humanoid Robot Parts Market? This Company Took the Lead," we took note of ENCOS Intelligence. This company has not been in business for very long. The team initially focused on developing robot joint modules, but its technology is highly distinctive, with final products characterized by high peak torque density and light weight. For example, one product named EC-A8112-P1-18 has a peak torque of 90 Nm and weighs 840 g, while another named EC-A4310-P2-36 has a peak torque of 36 Nm and weighs only 377 g.

On the EC-Hunter80-V01 open-source bipedal robot, we likewise see the lightweight feature. Unlike the humanoid robot bodies on the market today that often weigh tens of kilograms, the EC-Hunter80-V01 body weighs just over ten kilograms. For humanoid robots, this greatly increases the convenience and safety of development and indirectly improves the robustness of robot motion control.
This is because the joints used in the robot are precisely the two most competitive lightweight products from ENCOS Intelligent Technology: EC-A8112-P1-18 and EC-A4310-P2-36. The combination of these two self-developed and self-produced joints not only simplifies the design structure of the humanoid robot and reduces maintenance difficulty, but also reduces wiring harnesses and eliminates the time needed to integrate products from different sources, resulting in very significant overall hardware and software stability. In terms of performance, the peak torque density of both products exceeds 100 Nm/kg, providing strong power for the locomotion of a bipedal humanoid robot.

According to Robotics Lecture, previously ENCOS Intelligence focused mainly on module R&D and production and had not started humanoid robot research. However, in serving new customers, they often needed to invest considerable effort in communicating with clients and proving the performance of their modules on bipedal robots. Therefore, building a robot demonstration unit undoubtedly saves a great deal of time.
In the process of ecosystem building, ENCOS Intelligence fortunately met BridgeDP, a partner focused on robot algorithms. The two companies hit it off immediately and launched an ecosystem cooperation project, leveraging their respective strengths to rapidly create this open-source bipedal robot in order to better serve customers.
It is reported that BridgeDP is a startup composed of members with bachelor's, master's, and doctoral degrees from domestic and international universities such as SUSTech, Beihang University, HUST, and Tsinghua. The company is dedicated to developing motion control technology and modular products for humanoid and quadruped robots. Within six months of its founding, the team achieved a breakthrough in stable walking control of a humanoid bipedal robot from zero to one in just one month, thanks to its outstanding R&D capabilities. Next, the team will continue to invest in cutting-edge directions such as robot reinforcement learning and imitation learning, as well as the motion "cerebellum" module for humanoid robots.
In the cooperation between the two parties,ENCOS Intelligence provides the robot hardware and low-level communication solutions, while BridgeDP develops the adapted control algorithms. Subsequently, ENCOS Intelligence will be responsible for upgrading the corresponding modules.
ENCOS Intelligence believes that this robot is only a beginning, not an end. In the future, the two companies will cooperate long-term to continuously update and iterate this open-source version of the robot. In the R&D plan, for the future 2.0 version of the robot, ENCOS Intelligence hopes to add dual arms on the basis of the 1.0 version; the 3.0 version will add vision; and versions 3.0 and above will consider scaling the robot to around 1.7 m, making the robot more human-like and enabling it to perform more demos, thereby making robot research easier for developers.

Conclusion and Future
Previously,most open-source projects in the global robotics field were initiated abroad. As a rare Chinese open-source humanoid robot, the EC-Hunter80-V01 not only demonstrates the tremendous progress of China's current humanoid robot industry ecosystem but also showcases China's spirit of sharing and vision for deep exchange in the humanoid robot field. This open-source bipedal robot is believed to have the potential to accelerate the overall development of the industry and speed up the deployment of general-purpose humanoid robots worldwide.
"Global robot open source is inevitable sooner or later. Just as computer operating systems include both proprietary and open-source systems, and mobile phone systems also have open-source options, the future robot market will be enormous—there will certainly be open-source robots," said Zhu Zonghuang, CEO of ENCOS Intelligence. Shang Yangxing, founder of BridgeDP, added, "The development of the robotics community cannot be separated from the driving force of open source. We have benefited greatly from the open-source community in the past. We hope to continue contributing to the robotics community and, together with peers, make our own efforts for the development of general-purpose robots in China and globally."





界面新闻
深视新闻
深圳特区报
财新
36氪
钛媒体
中国新闻周刊
蓝鲸新闻
Z Finance
投资家网
雷峰网
亿欧
环球网
南方日报
极客公园



