EVOLVINGROBOT is a European Union (EU)-funded research project which has developed an artificial intelligence system to control tiny robots, enabling them to replicate the ‘swarming’ behaviour seen in insects such as bees or ants, or even in birds and fish. It is an innovation which could have far-reaching implications for a range of human activities, from medical to industrial, military and disaster relief.
© Peter Galbraith fotolia
“The breakthrough, allowing robots to act collectively in a way never achieved before, could herald a ‘paradigm change’ in robotics,” according to the project leader, Dr Roderich Gross of the Natural Robotics Lab at Sheffield University in the United Kingdom.
“The project addressed one of the grand challenges in robotics – how a robotic system can exhibit properties of living beings,” explains Dr Gross, whose work was supported by a Marie Curie European Reintegration Grant (ERG). The project succeeded in getting miniature robots, acting autonomously on the basis of very simple pre-programming, to perform a number of collective tasks. These tasks included gathering together in a single place, segregating themselves into distinct sub-groups, and cooperatively transporting objects.
The most important aspect was that these tasks required only minimal computing power. None of them required the robots to communicate with each other, and some did not require the robots to have memory or be able to compute at all. For example, to move an object, the robots were simply programmed to position themselves on the side of the object which obscured their view of the target area they were aiming for. Without communicating, the robots were thus able to ‘swarm’ in the right place to begin cooperatively pushing the object in the necessary direction.
“These results may lead to a paradigm change in robotics,” says Dr Gross. “Rather than building robots of increasing complexity, the results suggest that a range of capabilities could be realised with exceedingly simple mechanisms. Our truly minimalistic approach may pave the way for implementing massively distributed robotic systems at scales where conventional approaches to sensing and information processing are no longer applicable, for example, at the nanoscale. The amount of computing that can be used at such a scale is next to nothing.”
Enabling this technology at the nanoscale would be of major interest to the field of micro-medicine, where ‘nanobots’ are seen as the key to non-invasive treatment. The robots ability to gather together autonomously means individual robots could be injected into the body and then ‘self-assemble’ into the necessary larger groupings needed to carry out specific tasks.
“The idea is to make a robot so simple in terms of intelligence required that it could be further miniaturised, hopefully to a few microns in the next five to ten years – the size of a red blood cell. These robots could be used to transport drugs around the body or carry out treatments such as clearing blockages in the vascular system,” explains Dr Gross.
The technology also has exciting implications for manufacturing, based on the same idea of robotic self-assembly. “If you want to make a machine with dimensions of less than a millimetre, how can you have manufacturing equipment that is precise enough to do that?” asks Dr Gross. “In the future,” he says, “the goods might manufacture themselves, using robot ‘modules’ programmed to replicate a given template.”
At a larger scale, ‘swarming’ robots could be used in situations too dangerous or impractical for humans, such as search and rescue or military situations.
For development at the nanoscale, one issue that would need to be addressed is how to power the robots, since batteries are not possible. EVOLVINGROBOT has been examining ways in which the robots can harvest energy from their environment – another important property of living beings. The project has developed solar-powered robots, but other ways such as magnetic induction may also prove fruitful in the future.
For Dr Gross, the exciting significance of EVOLVINGROBOT is clear. “Less is more,” he says. “We have shown that robots can use very limited information and yet achieve tasks that, until now, no one knew could be done without computation. That is the most striking result.”