Retouching the droid

Researchers working on Europe’s most advanced humanoid robot hope to bring its intelligence up to the standard of its Japanese rivals. Jon Excell reports

While it doesn’t yet have the dancing skills of its Japanese rivals, iCub — billed as Europe’s most advanced humanoid robot — is catching up fast.

The toddler-sized robot — which made its debut appearance in the UK last month — is the centrepiece of the five-year £7.5m RobotCub programme, an EU-funded effort to advance knowledge and understanding in fields ranging from child psychology to robot cognition.

The project’s leader, Professor Darwin Caldwell, research director at the Italian Institute of Technology in Genoa, believes that mechanically the robot is up there with anything else around. ‘Our robot is mechanically comparable to the Japanese robots,’ he said. ‘It has a high number of degrees of freedom and is compact and powerful. The current model has 53 degrees of freedom and later in the year it will have around 70. By the end of the year we should have a fully compliant humanoid.’

Where the metre-tall droid lags behind its Asian rivals, he said, is in the onboard intelligence and control that would enable it to use its physical attributes; hence it can crawl but cannot yet walk. ‘The Japanese have had 10 years to develop the general control of software. They can walk, get towards running and do very clever dances. The programming they’ve done is tremendous and it makes the robot do something really interesting. Ours sits down and moves its limbs but it doesn’t do the really interesting stuff and that’s where we’re behind,’ added Caldwell.

He believes that the open nature of what is effectively a publicly funded academic project could soon change this. And the project is entering an interesting phase.

The last couple of years have been spent on the painstaking engineering needed to build it. Now this has been achieved, a small production line has been set up and by the end of the year around 20 different universities — including Imperial College, Plymouth and Manchester — will have their very own iCub.

By engaging the collective brains of Europe’s foremost robotics experts in this way, Caldwell hopes that advances will begin to come thick and fast. ‘The hope is that because there will be a significant number of these robots and a significant number of research groups programming them, we can hopefully build up that knowledge base very quickly. Very, very quickly we should be able to build up walking, walking quickly, walking sideways, even dancing. It was difficult to start doing this until the robot was built.’

While Caldwell certainly doesn’t discourage comparisons with commercial Japanese robots, iCub has a different purpose. While the efforts of Honda and Toyota are geared towards developing a commercial product, iCub is an EU-funded research project. ‘Our motivation in the RobotCub project has to be based around what research we can do that then spins off from this,’ added Caldwell.

The openness of the RobotCub project stands in stark contrast to the secretive nature of Japan’s robotics programmes. ‘Some of our guys have worked on HRP-2 [a humanoid robot under development by Tokyo’s Kawada Industries] and they still don’t know what the specifications are,’ said Caldwell. ‘It’s almost impossible to find out what torque the motors produce, what sort of frequency they move at. You can’t even go inside the software and do it.’

The reason for this secrecy, he added, is that Japanese industry seriously views humanoid robots as the viable basis of an industry that could one day rival automotive in scale. Robots such as Asimo are, he said, much more than wonderfully effective marketing tools. ‘Japanese firms are not thinking in a timescale that any European company would think of. Japan has a rapidly ageing population and by 2030 it will have the oldest population in the world. The Japanese see these robots — rightly or wrongly — as a way of taking care of people in 30 years’ time.’

It may sound a bit far-fetched, but with economies of scale, Caldwell believes the auto-industry analogy could be a valid one. ‘There are fewer components in a humanoid robot than in a car. It may currently cost hundreds of thousands but with mass production it could be realistic to get prices down to be comparable with those of cars.’

And although the European emphasis is different, Caldwell hasn’t ruled out the possibility of iCub evolving down a similar route. ‘It’s not impossible that the iCub in 20 years’ time couldn’t become one of these types of robot.’

Before that happens though, some huge technological leaps are required. For while today’s humanoid robots are mechanically sophisticated, their brain-power lags a long way behind science-fiction’s bold predictions.

‘Intelligence is where they currently fall down. With the hardware you occasionally get a big step-change, but there’s never been that in the area of intelligence,’ added Caldwell.

He hopes that the collaborative nature of RobotCub could help change this: ‘As more robots are brought online you’ll hopefully get a very rapid rise in people writing cognitive software for the iCub that they can share.’

And as robots become more intelligent, the business case for domestic androids may just gain credibility. ‘Some people claim that home automation technology could dispense with the need for domestic androids,’ said Caldwell. ‘That’s great if you’re building new houses, but most people aren’t going to be living in newly built houses. If you live in a Victorian house are you going to adapt it or leave so that only people can operate in it? If only people can operate in it, maybe you’re forced into a situation where a humanoid robot becomes an attractive option. If a robot is £10,000, that’s a lot less than putting someone in a nursing home. You can see why the Japanese think there might be a market for this.’