A digital twin system that could improve collaborative assembly of human-robot products

Robotic systems have already been introduced in many real-world settings, including some industrial and manufacturing facilities. In these facilities, robots can assist human workers on assembly lines and warehouses, assembling certain parts of products with high precision and then handing them over to human agents to perform additional actions.

In recent years, roboticists and computer scientists have attempted to develop increasingly advanced systems that could improve these interactions between robots and humans in industrial settings. Some proposed solutions rely on so-called “digital twin” systems, virtual models designed to accurately reproduce a physical object, such as specific products or components being manufactured.

Researchers at Nanjing University of Aeronautics and Astronautics in China recently introduced a new digital twin system that could improve collaboration between human and robotic agents in manufacturing environments. This system, presented in an article published in Robotics and computer integrated manufacturingcan create a virtual map of real-world environments to plan and execute appropriate robot behaviors when cooperating with humans on a given task.

“In industrial environments, current methods of building a human digital twin model rely on motion capture devices that require personnel to wear bulky equipment, which goes against the principle of flexible interaction advocated for the HRC,” wrote Zequn Zhang, Yuchen Ji and their colleagues. in their newspaper.

“Additionally, current methods do not model humans and robots in a unified space, which is both unintuitive and impractical for perceiving and understanding the overall environment. To address these limitations, this paper proposes a digital twin system for HRC.”

The digital twin system created by Zhang, Ji and their colleagues creates a virtual replica of a scene in which a human agent and a robot collaborate. Subsequently, he plans effective collaboration strategies and executes them in a real environment.

Previously proposed digital twin systems that rely on data collected by motion capture sensors have been found to sometimes achieve unsatisfactory results in the presence of occlusions (i.e., when objects or objects agents of interest are outside the field of view of the sensors or are hidden behind obstacles). ). The researchers thus developed a human mesh recovery algorithm, a computer technique that can help reconstruct occluded human bodies.

Additionally, Zhang, Ji and their colleagues introduced an uncertainty estimation technique into their system. This technique allows them to improve the performance of the action recognition algorithm, a component of their system trained to recognize different human actions, by controlling the risk that this algorithm makes errors.

The researchers evaluated their new digital twin system in a series of laboratory experiments, using a robot designed for deployment in an industrial setting. Their system was found to improve collaboration between this robot and a human agent in various tasks, including tasks involving polishing, picking up, assembling and putting down objects.

“The experimental results demonstrate the superiority of the proposed methods compared to the baseline methods,” said Zhang, Ji and their colleagues. “Finally, the feasibility and effectiveness of the HRC system are validated by a case study involving component assembly.”

The digital twin system developed by Zhang, Ji and their colleagues could soon be implemented on other robots for industrial use and tested in more detail in additional experiments. Ultimately, it could be introduced into real contexts, to improve collaboration between robots and humans. on various manufacturing and industrial tasks.

© 2023 Science X Network