Lego Blocks for Tomorrow’s Factories
IAS Visiting Fellow Frank Park on the Quest to Revolutionize Manufacturing
As some may be old enough to remember, Hong Kong in the 1970s was a thriving manufacturing hub, assembling everything from toys and textiles to consumer electronics. These and many other products are now manufactured in neighboring Shenzhen and Dongguan. Today, the Pearl River Delta region is indisputably the center of the world’s computer, communications, and consumer electronics (3C) industry. Few would welcome the return of such jobs to Hong Kong, however, as they are usually low-paid and menial, and sometimes even dangerous. Surprisingly, the process of assembling 3C products has changed very little since these jobs migrated north, with workers still responsible for the bulk of the assembly work.
Challenges on 3C Assembly Automation
Why have machines not displaced workers in 3C factories as they have in automobile factories? The answer, according to Prof Frank Park, IAS Visiting Fellow and Professor of Mechanical & Aerospace Engineering at Seoul National University, is that 3C assembly is deceptively difficult. “Today’s industrial robots are very good at welding, painting, and picking up objects and moving them from one place to another,” explained Park. “But machines are still no match for humans in assembling small flexible parts with varying shapes in tight spaces, threading, winding, and connecting wires and cables, visually distinguishing dust from a scratch, or detecting the subtle differences in the sound of a defective motor—all tasks that take humans just a few seconds to complete.”
Yet robotics is making steady progress. Robots can now grasp objects of varying shapes with some competence and assemble simple parts, at least in structured settings. Park and his HKUST Robotics Institute colleagues, Prof Michael Wang and Prof Zexiang Li, study the fundamental theoretical foundations of assembly processes as well as their realization in hardware and software.
“The notion that all assembly tasks can be reduced to a core set of primitive motions arranged in the right sequence, and that a similar core set of hardware devices can physically carry out these assembly sequences when paired with the right algorithms, has been around for at least 25 years,” said Park. “What has advanced considerably is our understanding of the mechanics and control of manipulation. With recent breakthroughs in mechatronics, materials, and computing, the implementation of these ideas is now very close to becoming reliable and cost-effective enough for industry use.”
A Solution to Automation
Two extremes have emerged in the current landscape of assembly automation. At one end lies sophisticated, flexible, all-purpose robots, which are expensive and difficult to program. At the other end lies highly specialized and efficient assembly machines that are mini-factories in themselves; these are expensive, arduous to design, and difficult to reconfigure for new and redesigned products.
Park and his colleagues have proposed a solution located between these two extremes: the use of simpler, less sophisticated robots driven by only one or two motors, augmented by soft smart grippers, modular fixtures, and customizable parts feeders, and paired with sophisticated software that allows these heterogeneous devices to be easily interconnected and controlled. He describes such robots as “Lego blocks” in the construction of assembly-automation systems, driven by an operating system and coordinated by higher-level application programs. The process is similar to putting together mobile-telephone hardware run by an Android or iOS operating system and coordinated by mobile applications.
However, many challenges must be met before these ideas can be realized in practice. How should hardware and algorithms be designed to enable flexible parts assembly? What can we do to make programming easier? How can we deploy minimalist assembly systems to ensure that they operate safely alongside human workers? Finding cost-effective solutions to these problems will require considerable effort. Some progress has already been made, however, Park and his Robotics Institute team recently developed an automated electrical motor inspection system based on deep learning, and are currently building soft grippers capable of manipulating flexible parts.
Park has been devoted to the study of robotics for over 20 years. His team has developed several mobile manipulation systems and simulations on robot design.
Hong Kong: A Global Hub for Robotics Research
If such automated-assembly technology can be successfully developed, the payoffs for Hong Kong and the region will be immeasurable. The struggles faced by today’s 3C companies due to rising wages and a shortage of workers are well known, and this confluence of economic and demographic factors has made manufacturing automation more important than ever. Park noted that “with the emergence of global companies like DJI, an industry leader in civilian drones founded by an HKUST alumnus, Hong Kong’s strong tradition in robotics research has begun to be recognized.”
Given this tradition and Hong Kong’s proximity to the “world’s factory” in Shenzhen and Dongguan, the city is ideally positioned to become a global innovation center for manufacturing automation. The manufacturing centers of the future will be small, locally distributed, fully automated, and interconnected, in stark contrast with the large factories employing hundreds of thousands of workers that we see today. “It is no secret that the best research is usually driven by local needs,” said Park. “It is only fitting that HKUST and the robotics research community in Hong Kong take the lead in these efforts and swing for the fences with high-risk, high-payoff projects that have the potential to transform society.