Making robots nearly as nimble as human hands, with tech powered by AI

Associate Professor Pham Quang Cuong (right) and Dr Hung Pham are co-founders of Nanyang Technology University's spin-off Eureka Robotics. PHOTO: EUREKA ROBOTICS

SINGAPORE - Robots can now manipulate tiny objects - including electronics components and optical lenses - with almost the same nimbleness as human hands, with the aid of new technology powered by artificial intelligence.

This allows for a level of precision handling only human workers were capable of until now, such as assembling engine gears.

The technology was developed by Eureka Robotics, a spin-off from Nanyang Technological University (NTU), and is currently intended for use in the software of industrial robots sold by tech firm Denso Wave, which is part of the Toyota Group.

With it, workers can be freed from mundane and repetitive tasks while ensuring high quality control by minimising human errors, said NTU, Eureka Robotics and Denso Wave in a statement on Nov 29.

The technology allows robots to possess both high accuracy and agility - traits that most industrial robots currently do not have.

A robot with high accuracy and low agility will be able to perform a task repeatedly with little error but is unable to vary its movements to allow for a wider scope of function.

An example would be the welding machines in an automotive factory, which can work only on identical car models placed at the same position each time.

Conversely, a robot with low accuracy and high agility will be able to perform a wider scope of tasks but is unable to do precise work.

One example would be a warehouse robot handling parcels of different sizes and in varying positions.

Associate Professor Pham Quang Cuong, who is Eureka Robotics' co-founder, told The Straits Times that three factors are needed for a robot to achieve high accuracy and agility.

The first is vision, he said. "For example, if objects are randomly placed, (the robot) needs to be able to find the position of an object and... go to the specific position."

The second is robot motion planning. In the same example, this means that the robot has to be able to adapt to handling various shapes and positions of objects.

The last factor is force control, which determines how much force the robot should use when performing its tasks. This is especially important in situations where high accuracy is required, such as in the assembly of objects.

Prof Pham said robots generally require several days of fine-tuning in order to achieve the correct force control needed.

But the software powered by the new technology requires only a single parameter to be set - the expected stiffness of contact - thereby allowing the required force control to be programmed into a robot in a shorter period.

The technology was first developed by Prof Pham and his Eureka Robotics co-founder Hung Pham in 2018. It was further refined before the two men published their research on it last year.

Prof Pham (right) said robots generally require several days of fine-tuning in order to achieve the correct force control needed. PHOTO: EUREKA ROBOTICS

The new software will be available as an optional update to the operating systems of Denso Wave's robots starting from this month.

The firm's clients will also have a choice to include the new technology in the software of new robots they purchase.

Mr Hiroyasu Baba, project manager in Denso Wave's factory automation/robotics business unit's product planning department, said the new technology is currently intended for use in the electronics and automotive industries.

The firm is exploring plans to implement it on robots handling and processing raw food, he said.

"Manufacturing is our main target," said Mr Takeshi Idomoto, who is assistant manager in the same department, adding that there are also plans to use the new technology in the agriculture industry.

As for use in households, this might be possible in the future, he said.

Join ST's Telegram channel and get the latest breaking news delivered to you.