Swarm-bots could boldly go where no man has gone before

Imitating insects such as ants, highly mobile small robots can accomplish physical tasks that no individual robot of the same size could manage. But if more sophisticated versions appear, then such machines could complete coordinated tasks in a way that could revolutionise the way we think about our world today.

An IST project in the Future and Emerging Technologies programme called Swarm-bots has lead the field in this area, and as a result the project partners have gained a great deal of publicity for their work. Marco Dorigo of the Université Libre de Bruxelles coordinated the project, and explains what the team achieved.

“We produced thirty-five complete s-bots [the individual bots that make up one swarm-bot], and completed many experiments with them,” he says. Just 12 cm in diameter, these mini-robots are packed with computing power, sensors and actuators, he adds.

In one trial, the s-bots linked up to bridge and thus pass over a hole in the ground. In another they jointly carried objects too weighty for a single robot to handle.

Dorigo likens the job of cooperatively finding objects to ant behaviour, although ants of course create their trails using pheromones.

He continues, “In our most complex experiment, we placed 20 robots in a big room to retrieve one object to a ‘nest’. This involved building a chain of eight robots spaced thirty centimetres apart and visible to one another. The other robots followed the chain to find and retrieve the object, all in just ten minutes.”

Each robot carries sophisticated technology, including a panoramic camera, sensors that detect sound, infrared, light, temperature and humidity, motors for the grippers (or claws), and WiFi and USB connections. A unique feature of the s-bots is their ability to attach to one another using the grippers – researchers have tested both hard and flexible versions.

The project partners successfully created s-bot types ranging from single autonomous machines to larger ‘swarms’. They also tested robotic control for simple tasks. “Control was vital to the project,” says Dorigo. “The robots autonomously attach to each other and move around in coordination. Their tracks and wheels guide their directional movement. Though the robots do not talk among themselves, they receive low-level signals – such as individual push and pull forces – allowing coordinated group movement.”

In one experiment, three robots called on a control program or ‘controller’ to minimise the traction force. Wirelessly downloaded into the robots, the controller was not designed by human beings.

The Swarm-bots coordinator won a Marie Curie Excellence Award in 2003 for his research on ant-colony optimisation and ant algorithms. In 2005, he was also awarded Belgium’s FNRS Prize (Prix Dr A. De Leeuw-Damry-Bourlart) for his contributions to the foundation of the ‘swarm intelligence’ discipline.

What could these robots do if they move beyond prototypes? “They could help in disasters, for instance picking through earthquake rubble to find trapped people. They could also be built into homes and buildings, emerging when needed to check out the local environment by creeping along the floor, walls or ceilings,” he says.

The s-bots are still far from real-world applications. Yet they have caught the attention of NASA. Famous for its Mars lander exploits, that organisation is seriously interested in robotic technologies that could be used to construct structures on other planets, if necessary without external direction.

Dorigo says challenges lie ahead in programming and controlling robots like this: “Today they exhibit simple or reactive behaviour. We would like to do things such as adaptive task allocation, for example using only ten robots out of one hundred to solve a problem, rather than all of them together.”

Dorigo's wish list also includes using muscle-like materials, as found in nature, instead of rigid metal and plastic. Enhanced memory would also help, allowing machines to find their own solutions to problems after assessing their environment.

Some of these wishes may be addressed in Swarmanoid, a three-year follow-on project launched in October 2006. Dorigo, who is also coordinating this project, hopes to create three different kinds of advanced robots. Featuring open-source software, these super s-bots will crawl, climb or fly, working alone or together.