Chinese scientists build factory robot that can read minds on the assembly line
- Trained robot monitored co-worker’s brain and muscle signals to predict needs, China Three Gorges University team says in domestic peer-reviewed paper
- China is in dire need of more powerful robotic technology to address its problems of a shrinking workforce and rising labour costs
The co-worker did not need to say or do anything when they needed a tool or a component, as the robot would recognise the intention almost instantly, picking up the object and putting it on the workstation, according to the developers.
“In modern industrial manufacturing, assembly work accounts for 45 per cent of the total workload, and 20-30 per cent of the total production cost,” project lead scientist Dong Yuanfa and his co-researchers said in a paper published in domestic peer-reviewed journal China Mechanical Engineering.
Collaborative robots, or “cobots”, could accelerate the pace of an assembly line, but their application remained limited because “their ability to recognise human intention is often inaccurate and unstable”, the paper said.
Humans and robots or autonomous machines have been working together in factories for decades, but are separated by fences in most places to avoid accidents.
In recent years, some advanced production plants such as car factories in Germany have introduced a fence-free work environment, with robots that swing into action only after a button is pressed. Such machines are equipped with safety sensors that stop them immediately if they come into physical contact with humans.
Some research teams have tried to build a new generation of “cobots” that can guess human intention by monitoring eye or body movements. However, these passive approaches suffered from problems like slow response and poor accuracy.
To overcome this, the robot created by Dong’s team was put through hundreds of hours of training by eight volunteers.
The volunteers were first asked to wear a non-invasive brain wave detector and the team found the robot could estimate their intent with just about 70 per cent accuracy.
However, the brain signal was quite weak. For the robot to get a clear message, the volunteer would need to concentrate very hard on the work at hand. But most of them were distracted by other thoughts after working on the repetitive assembly job for a while, the researchers said.
In contrast, the muscle signals, collected by a few sensors stuck to an arm, were more stable. Even though these too waned as the volunteer grew tired, a combination of both brain and muscle signals could help the robot estimate the worker’s next move in a second with unprecedented accuracy, according to the team.
However, it was unclear whether these laboratory results could be replicated in a real-life factory setting. The researchers could not be reached for comment by the time of publication.
According to the paper, there would be some challenges to the application of the new technology in a real factory setting. Though the brain and muscle detectors could be placed inside a worker’s cap and uniform, the quality of data could be affected by sweat or irregular movements.
But these problems could be overcome by feeding the robot with motion and visual data, the researchers suggested.
The number of industrial robots in China has been growing at a pace of 15 per cent annually since 2016, Wang Weiming, director of industrial equipment with the Ministry of Industry and Information Technology, said in Beijing on Tuesday.
There were 246 robots for every 10,000 workers in China, or twice the world average, Wang said, but the majority were built with technology developed in the West that sometimes could not cope with the challenging environments in China.
Hence, China is in desperate need of more powerful robotic technology to overcome problems such as a shrinking labour force amid sharply declining birth rates, and rising labour costs. By 2025, more than 70 per cent of large-scale factories in China would be using robots, he added.
Song Xiaogang, secretary general of the China Robot Industry Alliance, said the development of cobots would be a priority.
“There would be a leap from fence operation to robot-human collaboration,” he told the same press conference.
Some Chinese factories have asked workers to wear brain-reading helmets or use AI-controlled cameras to monitor their facial expressions.
Though the purpose was to detect fatigue, depression or other mental signs that might affect work efficiency or safety, critics have warned about privacy concerns.
In the United States, the use of robots has reduced industrial workers’ salaries, according to a study by the American Economic Association in 2019. Some studies by Chinese researchers in major economic zones such as the Pearl River Delta have found a similar phenomenon.
However, a study by Peking University researchers last month indicated that the mass application of robots in nearly 300 cities from 2006 to 2016 had helped Chinese factories produce more premium products of higher quality and increased the nation’s global competitiveness as a whole.