October 30, 2024 feature
Editors' notes
This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:
fact-checked
preprint
trusted source
proofread
Study conceptualizes GenAI-driven Industry 6.0 with a successful swarm demonstration
Since the industrial revolution, manufacturing processes have continuously evolved in alignment with technological advances. Recent innovations, particularly in the field of robotics, 3D printing and machine learning, could soon facilitate further change, potentially establishing a new generation for industry standards.
Researchers at Skoltech's Intelligent Robotics Laboratory recently introduced their vision for the future generation of industry, which they dubbed Industry 6.0. Their vision, outlined in a paper posted to the arXiv preprint server, was tested in an initial demonstration using real computational and robotic systems.
"We were inspired by our recent achievements in generative AI and its application in Robotics," Dzmitry Tsetserukou, supervising author for the study, told Tech Xplore. "This year we have developed several breakthrough systems, where GenAI helped to solve the issues that previously demanded a lot of human coding resources. It made me rethink the classical understanding of how things are manufactured that resulted in the Industry 6.0 concept."
While many associate the term "industrial revolution" with the first industrial revolution, which was marked by the advent of the steam engine, there have since been various other milestones marking further revolutions in industry. The second industrial revolution, which took place between 1870 and 1914, was fueled by the discovery of electricity and the chemical synthesis of new materials, which led to the introduction of assembly lines and mass production.
The third industrial revolution, on the other hand, has now been delineated as occurring between the 1960s and the 2000s, with the development of the first computers, robots, electronic devices and eventually the internet. These advances opened new possibilities for communication and enabled the automation of various tasks within industrial settings.
"The Fourth Industrial Revolution, or Industry 4.0. It is the concept proposed by Founder and President of World Economic Forum, Professor Klaus Schwab in 2016," Tsetserukou said. "In a nutshell, Industry 4.0 states for smart manufacturing. There are four foundational types of disruptive technologies to make it possible: IoT (connectivity of robots, 5G, digital twins), intelligence (machine learning), autonomous mobile robots (AMR), and additive manufacturing (3D printers, etc.)."
Theorists believe that we have now started to shift towards a fifth generation of Industry (i.e., Industry 5.0), in which humans and machines will be closely collaborating within the same environment. This new phase is associated with the development of collaborative robots, digital twins, deep learning techniques and other recently developed tools that enhance the co-operation between human users and machines.
While machines and computational tools are expected to play a greater role at this stage, cognitively demanding and decision-making tasks should still be executed by human experts. In computer science, this inclusion of humans in tasks is referred to as "human-in-the-loop."
"Industry 6.0, the concept I am proposing, leverages generative AI and a swarm of heterogeneous robots, where the human is out of the loop," explained Tsetserukou. "All stages, starting from blueprint design (CAD model) to the assembly are done autonomously.
"The only information that comes from a human is the idea for the product given through text, image or voice message, the output is the ready product (input-to-output model). Cloud generative AI (GenAI) supervises the design and manufacturing process, while each robot, equipped with AI, operates independently in a dynamic environment."
Essentially, Tsetserukou envisions a hierarchical machine-based workforce, in which cloud GenAI models guide the overall strategies executed by other computational models and robotic systems. The paper outlines his ideas for the future of industry, while also emphasizing its robustness and sustainability. In the fully automated scenario he describes, fabricated items are analyzed at every step of the manufacturing process, including assembly and product testing.
"When some malfunction is identified, Cloud GenAI finds the place when it happened and the reason of this. AI agents will supervise equipment repair and maintenance procedures. They will also be responsible for assessment of product quality, providing feedback on the product performance, discussion with the user whether their expectations are met," said Tsetserukou.
"The key 'workers' in new generation factories are swarms of heterogeneous cognitive robots. Industrial, collaborative, mobile, humanoid robots, drones along with AI-driven CNC machines and 3D printers are teaming together to achieve the optimal and prompt manufacturing of the designed product."
In his paper, Tsetserukou also forecasts the invention of a so-called flying conveyor. He suggests that instead of being assembled horizontally on tabletops or regular industrial conveyors, products could be manufactured vertically and carried to different stations by teams of flying robots.
"When Henry Ford invented the belt conveyor for assembly, it revolutionized mass production," explained Tsetserukou. "It also led to the horizontal layout of factories. I have proposed a flying conveyor where delivery of components can be done in any direction by a swarm of autonomous drones. It can potentially lead to more optimal processes, such as vertical layout of the factories, resembling vertical farming. The drones can team with autonomous mobile robots to handle a wide range of payloads."
In Tsetserukou's vision, flying robots would deliver items to a robotic cell. In this cell, they would be mechanically processed, after which they would brought back to the ground, where they would undergo quality checks and ultimately be stored.
"Industry 6.0 leverages GenAI-driven Digital Twins controlled (virtual AI factory)," said Tsetserukou. "We train robots in a digital world and transfer learning to a real world (Sim-To-Real). Industry 6.0 also suggests that each cognitive robot can teach other physical and virtual robots (Real-to-Sim-to-Real) and share collected datasets."
In addition to outlining his concept for Industry 6.0, Tsetserukou carried out an initial demonstration of its underlying technology, which was developed by 11 researchers at the ISR Lab led by Ph.D. student Artem Lykov, Dr. Miguel Altamirano Cabrera, and Ph.D. student Mikhail Konenkov.
"In the scenario, a human gives the command to design and manufacture a robotic gripper. LLM generates the executable Python code to get the object planar geometry," said Tsetserukou. "In the next stage, AI generates a 3D STL script for gripper components and sends it to a 3D printer. As soon as components are printed, the file with their positions and instructions for assembly is generated. Collaborative robot UR10 with Robotiq gripper grasps the handle of a metal plate with components from the 3D printer and loads it to the drone."
The researchers introduced magnetic materials both in the drones and the 3D printer, as this allowed them to detach the metal plate when needed. Using this plate, the drone could deliver components to a robotic assembly cell, where two UR3 robotic arms manipulated them, ultimately creating the final product. Notably, the behavior tree of all these co-bots was planned by a GenAI model.
"We measured the time required for CAD engineers to build the drawing and time spent for each stage of the process by humans and Industry 6.0," said Tsetserukou. "Surprisingly, we achieved a tremendous advantage with an average 4.4 factor of improvement. GenAI generated 3D CAD designs 47 times faster than its human counterparts, completing the task in 30 seconds."
In their initial demonstration, the researchers also found that drones could also deliver and assemble items faster than human agents. As robotic systems and computational tools are still improving, in future years these observed times could be reduced further.
"In addition to the robotic gripper, GenAI successfully designed a coffee grinder, glasses, scissors and even a quadcopter body," said Tsetserukou. "This does not mean that the design is perfect and optimal, but it proves that GenAI has a strong potential to learn how to deliver a high-quality design in the future. Thus, we can achieve a customized and unique product for each user instead of mass-production."
The recent paper by this team of researchers offers a possible forecast for the future of industry in which humans are no longer involved in the actual manufacturing of products. Tetserukou and his colleagues are now conducting additional studies exploring the role that emerging technologies could eventually play in industrial processes.
"If we scale up the idea, we can imagine that Industry 6.0 factories will collaborate with each other making and updating a global dataset, while sharing the most successful neural network architectures for the specific tasks to maintain a platform for industry-tailored GenAI architectures," added Tsetserukou.
"Our future research will be devoted to AI-driven optimization of processes in Industry 6.0, allowing the manufacturing of more sophisticated products. Most advanced VLM and LMM models will be embedded to improve the cognitive ability of the heterogeneous robotic swarm, so they can operate in a dynamic and cluttered environment."
More information: Artem Lykov et al, Industry 6.0: New Generation of Industry driven by Generative AI and Swarm of Heterogeneous Robots, arXiv (2024). DOI: 10.48550/arxiv.2409.10106
Journal information: arXiv
© 2024 Science X Network
Citation: Study conceptualizes GenAI-driven Industry 6.0 with a successful swarm demonstration (2024, October 30) retrieved 30 October 2024 from https://techxplore.com/news/2024-10-genai-driven-industry-successful-swarm.html This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.
Explore further
A deep reinforcement learning approach to enhance autonomous robotic grasping and assembly shares
Feedback to editors
