Researchers in Singapore are planning to develop a “smart” onsite 3D printer based on technology installed in two robots able to build flat pack furniture in record time.
The team at Nanyang Technological University (NTU) wowed reporters earlier this month when they used a 3D camera and a pair of industrial robots fitted with grippers and force sensors to build an Ikea chair in around 20 minutes, just 5 to 10 minutes longer than the task takes a human.
The machines were able to recognise the chair parts, strewn in different locations on the floor, work out how to pick them up, then fit them together without causing damage.
NTU now plans to exploit the same technology to augment the capabilities of a six-axis industrial robot it is developing to print concrete structures on construction sites.
Lead researcher Quang-Cuong Pham said: “The idea would be to use the technology we have developed to make 3D printing possible on construction sites, where conditions are less predictable.”
It could be over five years before the technology is deployed in the real world, as various technical (accuracy of localisation) and safety issues must be overcome, he added: “Our current technology can only deal with limited unpredictability – for greater unpredictability, especially being able to cope with humans moving around, much research is required before deployment on site.”
A more immediate impact is expected in offsite construction and the use of robots assembling prefabricated components in factories. NTU is currently working with manufacturing companies to develop robots to perform unpredictable tasks, such as grabbing, gluing or welding parts in different positions and directions, away from conventional assembly lines.
The robots used to build the £18 “Stefan” Ikea chair featured equipment already on the market, but sophisticated software, related to perception and motion, enables them to act with greater intelligence than regular industrial robots.
“Perception needed to precisely locate the chair parts, and motion planning are very important,” said assistant professor Pham. “With existing robotic systems you have to ‘teach’ the robot to perform a specific move, but here the chair part can be in any position and every time a different motion must be performed. The robot arm must also be able to avoid any obstacles and the other robot arm.”
The robots synthesise learning through a combination of taught instructions and complex algorithms. Engineers input a sequence of actions, translated from the Ikea assembly manual for the chair, the robots then work out the appropriate motions to accomplish those actions, based on what they see in the 3D camera.
“It’s like giving a child instructions to walk from one room to another in your apartment,” said Pham. “You might tell it to open the door, walk along the corridor, then enter the door to the next room. The child is able to work out how to achieve it by moving its feet one in front of the other and staying in balance.”
Innovation was also required for the interaction between the robots and the chair parts. The grippers incorporate force sensors that are able to “feel” the force between the wooden pin on one part and the surface of the other part to detect the hole and insert it.
Although the robots took just over 20 minutes to build the chair, over half of the time was spent planning moves – execution took just nine minutes.