Robots with a human touch

Date: 23/06/05

The creation of robots like humans is the ultimate technological challenge for scientists. Although often seen as Frankenstein-like children of scientists' pride or of the vivid imagination of science-fiction film directors, robots are starting to be considered as solutions to human problems. They are not just seen as time-saving domestic appliances, as sceptics forecast at the infant stage of robotics, but as tools that can understand and help humans.

The EU-funded RobotCup project, launched earlier this year, aims to understand human cognitive processes better by creating a humanoid robot intelligent enough to predict the future and act on its own will. A prototype will come out in mid-July. By the end of 2008, RobotCup, who will be the size of an average two-year-old child, will have nine brothers and sisters. 16 research centres are involved in the €8.5 million project; 11 of them are European, two American and three Japanese.

"In order to understand how something works, you have to be able to build it. That is the core of engineering science," says Giulio Sandini, professor of informatics at the University of Genoa and co-ordinator of the RobotCup project.

While previous attempts to create human-like machines have been focused on letting them carry out specific tasks, such as walking up and down stairs or playing football, scientists now want to add brainpower and will-power to robots.

In this new research area called 'neurobotics', technology needs help from other sciences. The building of intelligent robots brings together developmental psychology, computer science, information theory and neuro-physiology.

One of the participants in the RobotCup consortia is the department of psychology at Uppsala University in Sweden. The department has a 'baby-lab' where psychologists monitor babies to understand developmental psychology, focusing especially on the 'eye to hand' connection.

Studying how babies react to moving objects taught developmental psychologists how prediction is crucial for human motoric development. If a ball is rolling towards a baby, he or she instinctively predicts where the meeting point between the hand and the ball will be. The baby does not reach out for the point where he or she first spotted the ball, but in front of it.

A robot can have a functioning hand, be able to grab things with its fingers, but if it cannot predict the movement of a ball, it will never be able to grab it. The study on babies showed that movement is not just reaction.

Before the baby can reach out for a spotted object, the eye tells the brain to order the arm to move. The order goes via the nerve system and out to muscles. Finally the muscles react and move the arm exactly the way the brain intended, to grab the object.

Sandini explains how knowledge about the will-power steered mental process behind grabbing an object is crucial in designing RobotCup and also in the making of good prostheses.

If you have an arm prosthesis, your brain tells you to move your arm, but the prosthesis has to respond efficiently to the order. "To create better prostheses we have to study the interaction between eye, brain muscles and hand, and find a way to translate this information into technology," he says. This is one of the areas where robots can become useful study-objects.

Claes von Hofsten, who runs the baby-lab at Uppsala University, explains that the robots can also be useful also in psychology.

"By building a robot that mirrors children's behaviour, we hope to confirm theses we have on the mental processes behind learning. A baby cannot tell us that, as he or she does not speak. Also, there are ethical issues with testing human behaviour or how biological beings interact with actual humans. And above all, it takes time," he says.

Robots have also been found to be able to stimulate children with autism, inaccessible to human stimuli. If a robot that reflects and imitates the child is created, psychologists hope to be able to find patterns of autistic children's behaviour and draw conclusions on how to interact with them.

Sandini says the questions scientists ask about human behaviour have become more complex since they first started using robots to understand humans.

"Earlier scientists wanted to know how our brain steered the vision system and they created models to prove simple theories. Today's scientists want to know why our vision is attracted by certain colours, or why a child reacts negatively to, say, a jerky walking style. To make 'friendly' robots we need to know exactly what kind of behaviour or movements are consciously or sub-consciously perceived as friendly," he says.

Other scientists look upon robotics as mere fun. RoboCup is a robotic football contest that started in 1997 to highlight and promote robotics research and create opportunities for scientists to exchange information on technical achievements. Launched by a Japan-based organisation, together with various research centres around the world, the project aims at developing a team of fully autonomous humanoid robots that can win against the human world champion football team - in the year 2050.

As scientists put a friendly face on robots and try to make them more like humans, the study of babies has become a key to success.

• Primary Source