Called YuMi, it is the firm’s first ‘collaborative’ robot, designed to be safe, useful and accepted around humans.
The idea is that a YuMi can be set up to accurately perform repetitive manipulation operations on a production line or for batch manufacture, interacting with people where needed.
Sony UK already has one, performing pick-and-place on a line making cameras in Bridgend.
Collaborative is a big word in the world of robotics.
The types of robots that make cars, some of which can shift a tonne at high speed, are safe so long as people don’t wander into their working environment. To keep humans from harm in those settings the robots are caged and inaccessible while they operate.
Then there are robots surrounded by light beam ‘fences’, which can be trusted to do an emergency stop if someone gets too close. Others are intended to be handed parts by humans, and to slow down or pause when approached.
None of these is collaborative, according to ABB. To get that accolade, the robot has to safely share space with and freely interact with a human.
YuMi is designed to be inherently safe, according to Mike Wilson, marketing manager at ABB. “You don’t have to put guarding around it,” he said. To start with, it has padding and no places that anyone can trap a finger.
For active safety, rather than use cameras, light beams or pressure pads, ABB senses torque in YuMi’s joints.
In this way, the force any part of the arms can exert is limited to 25Nm, said Wilson, and the arms stops as soon as they meet unexpected torque. Gripper force is also limited to 25Nm.
As YuMi refuses to manipulate anything that weighs more than 0.5kg, the maximum blow it can deliver to anyone or anything unexpected is 0.5kg at its maximum gripper speed of 1.5m/s.
This does not mean everything that it can do is inherently safe, just that YuMi itself is not dangerous to work around. “The robot is UL-approved as safe, but if you put a knife in the gripper, you have to protect against the knife,” said Wilson.
Because it has two arms, it can pass things from gripper to gripper to pick an object up by an exposed face, for example, then place it down again on the formerly exposed face – something that is difficult for a one-armed robot.
With two arms, it can also assemble click-fix parts: one demonstration is of a YuMi aligning and assembling male and female parts of a bayonet connector using force sensing to detect the click of proper assembly.
Grippers can rotate continuously, so screws can be inserted and caps put on bottles. “There is an awful lot of interest for toy assembly,” said Wilson.
The robot also has multiple industry-standard busses to allow it to communicate with machines it may need to take parts from or pass parts to.
For picking up fine parts, alternative gripper tips can be 3D-printed, and there are optional suction ports for pick-up nozzles to handle surface-mount parts, for example. Optional cameras, even cameras inside the grippers, can be used for machine-vision feedback.
Position repeatability is 0.02mm – 20µm – good enough for most pick‑and‑place ativity. In passing, Wilson pointed out that the huge spot-welding robots used for car production are accurate to 0.01mm.
For convenience, YuMi is entirely self-contained: the arms have no exposed cables, all necessary controller electronics is built-in and software for tasks, including gripper vision, is internal.
Power comes from plugging into a normal mains socket (100V-240V) via a standard ‘IEC’ lead.
It is also physically small and portable: YuMi sits inside a 400x500mm rectangle and weighs 38kg complete.
Programming can be as simple as switching the machine into a ‘motion copy’ mode, taking hold of the grippers, and moving them through the range of motion expected – the servos are set to be ‘soft’ during this, which means they support the arm weight but offer only token physical resistance.
To go along with this there is a tablet‑based app to open and close the fingers and other detailed work.
When set to operate, the grippers and fingers follow the demonstrated route, although the arms will assume whatever angles and positions deemed best by the robot to achieve those finger and gripper positions. More sophisticatedly, the robot can be manually positioned into roughly the correct place, then fine final movements can be programmed through a keyboard.
On top of the tablet app, ABB has its own high-spec computer-based movement control development programme for off-line programming. “The software language is the same for all [ABB] robots, so it is easy to move to larger robots,” said Wilson.
This does not mean that creating a reliable set of actions is trivial.
Moving easily-gripped objects from one place to another is one thing, but identifying a misassembled part and discarding it part-way through an operation, or recovering after dropping a slippery part, requires programming sophistication and possibly camera feedback.
According to Wilson, while ABB’s week-long training course will not make everyone an expert in robotics, it will enable most people to programme a robot to be productive, including incorporating the use of camera-based feedback.
“We teach 900 people per year how to train robots, and these are not robot people, these are welders and production people,” said Wilson. “It can do the repetitive bits and it allows a worker to step up and become a robot trainer.”
ABB wants the robot to be accepted by production‑line employees, particularly as it is aimed at companies that have never had robots before.
YuMi is the size of a small human, its arms move like human arms, and somehow the firm has made a headless torso with arms look cute.
Robots and jobs
For Wilson part of making YuMi acceptable is persuading people that robots will not take their jobs – although he does admit the potential for a “short‑term decrease in employment” before the long‑term increase in employment he foresees.
He points out that the highly robotised UK car industry employs tens of thousands of people, and argues that the UK wouldn’t have a car industry at all without extensive use of robots.
Wilson wields a plot of increase in robot use between 1993 and 2007 against percentage change in manufacturing employment between 1996 and 2012 in several manufacturing countries, showing, he says, that “there is no relationship between a country’s use of robots and percentage of manufacturing jobs lost”.
According to ABB, Future-proofing UK manufacturing, a report by Barclays: “Even a moderate increase of £1.24bn in automation investment could raise the overall value added by the manufacturing sector to the UK economy by £60.5bn” and “would be expected to amount to 33,000 manufacturing jobs by 2020 and 73,000 jobs by 2025.”
From the UK’s manufacturing sector containing around 87,000 SMEs (each with less than 250 employees), ABB sees 10,000-20,000 companies as potential users for ‘Industry 4.0’ technologies, including robots.
YuMi was designed at ABB in Sweden and China, is being made in China, and will be made in Sweden.
Why the Chinese connection? “China is now installing more robots than anyone else in the world, even Germany,” said Wilson.
The price for YuMi is between £40,000 and £46,000, depending on camera and gripper options. It can be supplied without grippers so customers can build their own manipulators. Custom fingers for the grippers can be 3D printed to suit the parts they are handling.
ABB showed YuMi in a couple of demonstrations – quick demonstrations rather than production grade.
One was making Lego cars (see photos), using vision to locate parts spread randomly on trays.
And it made cars – making great job of locating, picking-up and orientating parts.
However, the grippers were not customised to manipulate Lego, so occasionally it failed to complete a step when the Lego slipped or fell as after pick-up, and there was no feedback yet programmed to detect this had occurred.
Also, Lego requires an unpredictable amount of force for assembly, sometimes above YuMi’s set limits.
The firm was also part-way through developing a programme for pulling an O-ring around a collar, requiring both arms simultaneously.
With no visual feedback at the time of demonstrating, the O-ring had to be in exactly in the right starting place for the action to be completed.
Your faithful reporter, who knows almost nothing about production robots observed:
- YuMi is very capable and can do tasks without visual feedback if parts to be manipulated are offered in an exact position and orientation.
- It can do tasks on parts presented without exact alignment if visual feedback is programmed.
- Visual feedback throughout a task allows recovery if slippage occurs after initial gripping and orientation.
- Custom fingers increase the chance of completing a task and reduce the need for visual feedback.
- Necessary weight and force limits preclude some applications.