The robot child
One method to truly understand how something works is to build a replica. It is an idea that has been in practice for centuries and is now being used by a team of European researchers wanting to understand how humans learn. The result is a small robot named iCub that will eventually be able to learn like a two-year-old child.
The iCub prototype has been developed over the last three years at the Italian Institute of Technology in Genova. The research robot has been designed to be able to interact with its environment. Its hands can manipulate objects and its eyes can move without having to turn its head. Part of its software, known as the attention system, processes the images “seen” by the robot and filters for things in its environment that could be of interest (for example, faces, moving hands or something that is very bright or colourful).
The European project, RobotCub, which has lead to the creation of iCub involves experts in neuroscience, psychology and robotics. They want to build a humanoid cognitive system in order to investigate the theory that the learning process in humans is the result of physical interaction with their environment. This is why the robots physical design and sophisticated hands are so important; to learn like a human it should also be able to interact like a human.
It was decided not to model iCub on an adult, so the engineers set out to copy the body and intellectual capabilities of a 2-year-old child. Asides from the fact that adult intelligence is extremely complicated, it is hoped that the robot will be able to develop and better itself over time through learning and problem solving.
Help on understanding how children learn came from Sweden. Claes von Hofsten is a professor in psychology at the Uppsala University and has spent 30 years studying infant development. His task in the RobotCub project was to create a so-called “cognitive roadmap” - showing the kinds of cognitive abilities the robot should develop and what abilities need to be built into the robot.
Claes von Hofsten established various developments of awareness, perception and reasoning in babies that will be required as basic features of iCub. For example, in one experiment the professor analysed the eye movements of a 6-month-old baby watching a film. This showed that the robot should be able to direct its attention and track moving objects. Another experiment, in which another 6-month-old baby learned to catch a toy, shows the robotic engineers that iCub must be able to make simple predictions.
Establishing these basic cognitive skills is no easy task for the robotics team in Genova. Giorgio Metta, robotics professor at the Italian Institute of Technology, cites two major problems facing them: firstly, the artificial vision and, secondly, the ability to learn from its mistakes.
Upon completion, the robot will be put into use as a research tool. There are already orders for iCubs from 19 European research institutions. It is also being used as a common and practical platform for open source software and hardware. But the iCub prototype that has been built is just the beginning of a scientific and technological challenge, leading to a better understanding of ourselves and how we learn.