Integrating physical and AI technologies such as force perception, vision, and deep learning, it also has versatility, intelligence, and the ability to complete complex tasks, and can achieve excellent industrial-grade performance in an uncertain working environment.
A Chinese team assembled by Stanford AI and Robotics Lab created the adaptive robotic arm Rizon (Dawn).
And this is also the world’s first adaptive robot, combined with force control and AI technology, which can lay the foundation for the final completion of complex tasks in manufacturing, medical, retail, and other fields through robots.
At present, this robot arm has been applied in the fields of 3C electronic product assembly, data center operation and maintenance operations, auto parts polishing, and wood product polishing.
What kind of robotic arm is this?
Adaptive robotic arm
Dawn is a 7-axis robotic arm that does not require precise positioning and is completed by the hand-eye cooperation of the embedded computer vision module.
In terms of force sensors, the joint design and control hardware has been specially optimized, and even the core hardware is self-developed to achieve the best force sense control ability.
In addition, there is also a multi-layer intelligent system built by a dedicated processor driven by AI and robot algorithms, which provides perception and force-guided control capabilities, extremely flexible task planning capabilities, and real-time adaptive capabilities.
At this year’s Hannover Industrial Exhibition site, the dawn robotic arm met various challenges.
Such as precise perception and recognition:
The above capabilities are also the core features of the third-generation adaptive robotic arm.
Different from the first-generation high-precision position control manipulators and the second-generation collaborative manipulators, the third-generation adaptive manipulators have three major characteristics, which make it a revolutionary difference from the first and second-generation manipulators:
The error tolerance is high. On the production line, errors may come from product tolerances, process errors or deformation under stress, errors accumulated in the assembly or inspection process, and position judgment errors of the AI vision system. This cannot be completely avoided. The adaptive manipulator can overcome these errors and ensure excellent working ability, and therefore it is more adaptable to uncertain production environments than any robot in the past.
Strong anti-interference. When the robot base (AGV) shakes, the tools assembled on the robot vibrate, or human employees touch and interfere, the third-generation adaptive manipulator can well offset or adapt to the interference to complete the task.
Strong ability to transfer work. The third-generation adaptive robotic arm has hierarchical intelligence based on force control and vision. It can handle a large number of similar but not identical tasks with simple configuration and solve the difficulties in the past production line, such as assemble some shapes through the same production line Parts or connectors with different and similar assembly methods.
So who is the creator of this world’s first adaptive robot?
This is an AI robotics company start-up company founded in 2016. The core founding team comes from the Stanford University Robotics and Artificial Intelligence Laboratory.
Several founding members have already gained fame in their respective fields.
CEO and Chief Product Officer Wang Shiquan graduated from Stanford University’s Bionic and Smart Operation Laboratory and Artificial Intelligence Laboratory and has more than 10 years of experience in robotics research and development.
During his Ph.D. period, Wang Shiquan focused on the robot body and dexterous hand design, grasping and dexterous operation theory and modeling simulation, humanoid robot sensor control in complex environments, artificial intelligence, and robot theory integration, etc.
Its leading research on humanoid rock climbing robot won the IROS Best Paper Award in 2016, the international top robotics conference, and its research projects also include the underwater humanoid robot Ocean One and the humanoid robot research SupraPed in complex environments. He has published more than 20 international top journals and conference papers (Science, IJRR, IROS, ICRA, RSS), and obtained a number of technical invention patents in the United States and China.
Chief robotics scientist Zhong Shuyun, Ph.D. graduated from National Taiwan University in mechanical engineering, systems, and control. He was a postdoctoral fellow in the Artificial Intelligence Laboratory of Stanford University and a visiting scholar at the Carnegie Mellon University Robotics Institute. He has more than 15 years of experience in robotics research and development.
Zhong Shuyun has published more than ten international top journals and conference papers (TRO, IROS, ICRA, etc.). His research on humanoid robots in complex environments (SupraPed) was published as annual key research by the National Science Foundation in 2015.
CTO Ye Xiyang, Ph.D. graduated from Stanford University Artificial Intelligence Laboratory, majoring in mechanical engineering, under the tutelage of Professor Oussama Khatib, founder of humanoid robots, chairman of the International Robot Research Foundation, and IEEE academician.
He used to be a researcher at the Singapore Science and Technology Research Agency. He has more than 10 years of experience in robotics research and development. He has participated in a number of US national robotics projects. In 2016, he led the underwater humanoid robot Ocean One successfully tested at a depth of 100 meters in the ocean and assisted in 350 The archaeological project of the shipwreck two years ago.
There is also the chief AI scientist Lu Cewu, who graduated with a Ph.D. in computer science from the Chinese University of Hong Kong and worked as a postdoctoral fellow in the Artificial Intelligence Laboratory of Stanford University.
Lu Cewu is also a researcher and doctoral supervisor of the Department of Computer Science of Shanghai Jiaotong University. He founded and led the Machine Vision and Intelligence Laboratory (MVIG) of Shanghai Jiaotong University. A total of more than 60 top journals such as CVPR/ICCV/PAMI were published (including reception)
The team led by him ranks first in the world in the human body posture MPII evaluation. The top computer vision conferences such as CVPR/ICCV received more than 60 articles from him. One article is the top graphics and image conference SIGGRAPH has the highest number of citations in the past five years.
Industrial application test
Of course, the dawn robotic arm built by the Feixi team is no longer a product in the research and development stage. It has already exerted its productivity value in the industrial field.
At the new energy vehicle manufacturer, the Fawn Robotic Arm has undertaken component plug insertion and removal and connection quality testing.
Dawn uses AI vision technology to accurately detect and grasp plugs with wires without the need for positioning tooling, and use force guidance to adjust the position and posture of the plugs in real-time to complete the insertion task with high quality. The work process is extremely demanding. Precise force control and computer vision technology.
The value can also be glimpsed from the actual effect comparison. Lexi robotic arms can complete all tasks in a workstation, replace human workers, and complete higher-quality tasks than humans through automated solutions without additional manual inspection.
In addition, there is a field of 3C consumer electronics.
The Feixi team has reformed the automated production line of assembly parts for OEM manufacturers and provided a one-stop solution with extremely high flexibility so that the production line can be completed for different models of smartphones without or with minimal adjustments. Task to help realize flexible production.
In terms of value, the OEM capital expenditure and total production expenditure can be reduced by more than 50% within three years.
And this is just a small test, Feixi team introduced, it can be applied in more scenarios and industries in the future.
First, complete complex tasks that were previously difficult to accomplish and broaden the possibilities of robots.
Based on advanced AI and robotics technology, Flexi enables the robotic arm to complete complex actions with the cooperation of hands and eyes like humans and optimizes corresponding actions or task strategies in the process of tasks through AI technology.
Second, it can be applied to different production lines.
Fei Xi has developed a hierarchical general AI platform that allows companies to easily allow robotic arms to transplant previous tasks into similar scenarios, learn similar objects, new skills, and quickly adjust to new production lines. Reduce the investment required for installation and commissioning.
And application expansion in different industries.
Feixi has created a one-stop solution that has the ability to effectively complete work under uncertain environments and is highly versatile.
In the future, it can be used not only in industrial manufacturing, logistics, agriculture, and other industries to improve production efficiency and profit margins but also in medical, retail, and other environments where there is more human interference.