Key Technologies for Developing Robots in the Era of Smart Manufacturing

Humanity has experienced the first to the third industrial revolutions, from the power of the steam engine to the electricity of the second industrial revolution, to the computing power of today’s information age. Against this backdrop, the era of intelligence has emerged, where intelligent robots play an important role in various fields such as industry, agriculture, and national defense. The appearance of robots predates the concept of artificial intelligence; the earliest mobile robot was the guide car invented in China during the Eastern Han Dynasty. The first industrial robot, the PUMA robot, was invented in 1978, marking the peak of development from industrial robots to special operation robots, service robots, and unmanned systems.

Robots play a crucial role in the main battlefields of the national economy. In recent years, the country has widely applied robots for processing, welding, polishing, measuring, spraying, and other tasks in large ships, rail transit equipment, aerospace equipment, and especially in new energy vehicles, making robots an essential tool for a manufacturing powerhouse.

Robots are also vital in industrial manufacturing, particularly in major infrastructure construction related to people’s livelihoods, as well as in dangerous and harsh environments. They have become indispensable tools in engineering, infrastructure, and major facilities.

In recent years, robots have been undergoing a transformation and upgrade in the manufacturing industry, with “machines replacing humans” addressing the pain points and difficulties faced by manufacturing enterprises. First, there is the aging workforce and difficulty in recruitment, especially after the COVID-19 pandemic; factories that are automated can resume operations quickly, while those that are not automated do so slowly. Second, as human expectations for products increase, with one product often requiring thousands of processes, many of which are difficult to complete manually within short cycles, robots play a significant role. Based on these points, robots have developed rapidly in recent years.

Various developed countries around the world have launched robot strategies, making robots a competitive high ground in global intelligent manufacturing. For example, the U.S. National Robotics Plan 2.0, Germany’s Industry 4.0 strategy, China’s intelligent manufacturing development strategy, and Japan’s new robot strategy.

The U.S. robot strategy focuses on developing collaborative and interactive robots, as well as manufacturing and military robots.

Germany emphasizes intelligent manufacturing, promoting the transformation and upgrading of manufacturing, and providing more robotic services for the elderly and disabled.

Japan’s strategic plan focuses on developing industrial robots, nursing robots, and medical rehabilitation robots.

Europe, as an agricultural developed region, concentrates on agricultural robots and social service robots.

China, on the other hand, is developing various types of robots across the board in manufacturing, service industries, special operations, exploration, aerospace, and unmanned systems.

In this context, the country has set high requirements for robots in intelligent manufacturing, artificial intelligence, and becoming a robotics powerhouse, making robots the carriers of application.

There are many definitions of robots; originally, robots were mechanized and automated devices. Broadly speaking, robots encompass mechanics, materials, electronics, automatic control, computers, network communication, and artificial intelligence, making it a comprehensive interdisciplinary field.

From the perspective of core technologies, robots primarily consist of environmental perception, the robot’s body structure, path planning, optimization scheduling, and control systems for learning and executing tasks.

Robots can be categorized by use into industrial, agricultural, medical, etc. Structurally, they can be divided into humanoid, bipedal, stationary, etc. Spatially, they can be categorized into land, air, etc.

In recent years, robot development has mainly focused on industrial manufacturing. First, 90% of robots are used in new energy vehicles and traditional vehicle manufacturing, such as unmanned operation robots playing important roles in spraying and assembly processes. Second, underwater robots, such as those for diving exploration and deep-sea detection. Third, space robots have become essential tools for exploring the moon, Mars, and outer space. Fourth, land robots primarily focus on special industries and unmanned driving.

Robots primarily include four key technologies: first, the perception system; second, the decision-making system; third, the execution system; fourth, motion control.

From a control perspective, it is a closed-loop feedback control system, from sensory recognition to decision planning, to control execution, and to human-machine interaction.

In recent years, the rapid development of robots has benefited from the interdisciplinary integration. Artificial intelligence has promoted the development of machines, and the advancement of machines has pushed forward artificial intelligence. There are many types of robot platforms, including industrial, flying, mobile, medical, marine, and space, with a wide range of application fields. They play significant roles and demonstrate effectiveness in manufacturing, logistics, healthcare, daily life, and marine environments.

Robots have made significant progress in aerospace manufacturing and assembly, smart agriculture, intelligent harvesting, space exploration, deep-sea exploration, and scientific research. Intelligent service robots are the future direction of development, with vast industrialization potential, whether in assisting the elderly and disabled, healthcare, or rehabilitation, requiring widespread application of robots.

In recent years, medical robots have developed notably, as they excel in precision in minimally invasive surgeries. Exoskeleton robots assist in the rehabilitation of disabled individuals, and brain-machine interface robots have played excellent roles in assisting the treatment and recovery of cerebral palsy. Furthermore, special industries, such as nuclear power, space station construction, special operation robots, and epidemic prevention robots, have also emerged.

The future development of robots and the challenges faced in the past decade can be summarized as focusing on new materials and power. Secondly, bionic robots, crowd imitation robots, swarm robots, and related fields such as navigation, social interaction, and healthcare are also important. Regardless of how development proceeds, the ultimate goal of robots is to evolve towards intelligence, autonomy, and collaboration.

What key technologies must be overcome to develop a robot? Aside from the robot’s body, first, the robot must possess perception capabilities for both environmental and self-awareness; second, as a mobile carrier, the robot must control its structure; third, during operations, robots need effective path planning and must enhance their intelligence level, as learning is the only way to improve robot capabilities; fourth, robots must make good decisions and control. These aspects reveal four major key technologies that robots must overcome: first, the robot carrier; second, robot perception; third, robot decision-making and planning; fourth, robot motion control.

Our team is currently researching robot binocular vision sensors. Through learning models from binocular vision sensors, we can estimate the environment’s objectives. In the past, we relied on extensive mathematical computations and image processing, but now we use deep learning to build the architecture. Once the learning model is established, we can gather complex backgrounds and scene targets to create a three-dimensional map for the robot. For instance, to enable a robot to effectively select a disorganized product, it must first acquire three-dimensional information about the target, perform feature and model analysis, and then conduct three-dimensional reconstruction before editing the execution command system. At the same time, a complete three-dimensional reconstruction map must be generated for the robot, providing it with three-dimensional environmental perception, so the robot can start working.

The second key technology for robots is developing the perception system, which allows robots to move. High-performance controller instruction systems derive from perceptual information, such as compliant control, visual servo control, intelligent control, and multi-robot collaborative control in manufacturing robots. The most basic control unit of a robot can be implemented through high-performance motor drive control. Once motor control is established, robots can grasp and identify objects, achieving tasks more accurately. Reinforcement learning addresses the decision-making problem in robot learning, while deep learning solves the perception problem in robots, making these two technologies extremely important.

Previously, fixed processing and manufacturing robots could only mechanically execute instructions, but reinforcement learning has made robots smarter, giving them life and soul; they can not only understand problems and recognize information but also distill a series of complex postures and operate autonomously.

The third key technology for robots is task execution. To solve the robot’s dexterous end-effector control system, two core issues must be addressed: first, kinematics; second, dynamics. By planning paths, we can direct the robot’s action trajectory to perform various complex tasks. For example, directing a robot to assemble large aerospace components, continuously iterating learning allows the robot to enter more scenarios and perform better. For instance, Boston Dynamics robots have transitioned from traditional electromechanical control systems to force-position hybrid control, elevating the perception system to machine learning. Through machine learning, they can effectively control gait and conduct scene recognition in various scenarios. This is based on deep learning three-dimensional perception to solve complex structural control systems.

The fourth is control, where operating large controllers can also employ this technology: first, coordinate planning; second, scheduling; third, motion control. The key to multi-robot collaboration, besides environmental perception, is also to solve collaborative perception and environmental cognition, enabling multiple robots to complete common tasks in an orderly manner. For example, in future battlefield swarm combat systems, addressing collaborative perception, collaborative planning, and collaborative control is essential.

The most mature application of robots remains in flexible manufacturing, where product updates equate to reconfiguration, updating software and downloading it to distributed control systems to achieve flexible control of production lines through multi-machine collaborative scheduling. First, focus on the body structure; second, focus on perception intelligence; third, focus on planning and decision intelligence; fourth, focus on intelligent control. Additionally, once robots are networked, they must also detect and eliminate viruses, then reconstruct control systems. Addressing these points can solve manufacturing, aerospace, logistics, healthcare, and household issues.

Our team has done some work in intelligent manufacturing.

To create intelligent manufacturing with robots, we must first establish a system architecture, which includes technologies for loading and unloading, welding, polishing, inspection, and transportation. The welding of new energy vehicles and ships that we have developed is all completed by robots. For instance, the production speed of medical masks and vaccines has benefited from robots. For vaccines, first, to avoid infection; second, robots are used for testing; third, robots are used for disassembly; fourth, robots are used for bottling; fifth, robots are used for packaging. In the beverage industry, robots have solved visual inspection and manual assembly tasks, achieving “robots replacing humans”.

Besides industrial manufacturing, robots play significant roles in major engineering projects, such as complex power maintenance, major infrastructure maintenance, large tunnels, and large infrastructure construction, including mining, petrochemicals, and metallurgy. In recent years, many robots have entered hazardous and harsh environments for operations.

Robots have also played essential roles in major epidemic prevention efforts. After the outbreak of the pandemic, various companies developed different robots, and high-end surgical robots have developed rapidly. Moreover, intelligent unmanned systems, such as aerial and underwater unmanned systems, have addressed applications in various complex scenarios, including urban, forest, and deep-sea combat. In recent years, countries have been developing unmanned combat systems, including military systems, such as air, land, and sea collaborative systems.

In recent years, robots have made rapid advancements, primarily enhancing autonomy.

Today’s robots are still in the 1.0 automation system or 2.0 digital system; the future requires embedding core artificial intelligence technologies into robot systems, namely intelligent collaborative robots, integrating artificial intelligence technologies into robots. For instance, cognitive intelligence technology and collaborative swarm intelligence are two key points, designing robots’ memory, perception, and action to resemble a human brain’s intelligent autonomous network control system, emphasizing the perceptual, behavioral, and cognitive intelligence of humanoid robots, enabling them to have the ability for autonomous recognition and operation.

Robots will play an even greater role in intelligent manufacturing, addressing small batch and customized production; the development of unmanned factories; unmanned driving; the development of intelligent networking; and human-robot collaborative robots.

Intelligent robots can be summarized as having several points in the deep integration of artificial intelligence and robotics technology: one is perception intelligence, one is cognitive intelligence, one is decision-making intelligence, and one is control intelligence.

Future development trends of robots can be summarized as follows: first, networking; second, autonomy; third, collaboration; fourth, dexterity.

The top-level design of robots must be planned and standardized, and a strong robot R&D team must be created to ensure the successful development of robots.

(This article is based on the author’s speech at the 2022 China Artificial Intelligence Conference (CCAI Changsha) main forum, and has not been reviewed by the author)

END