Robots are undoubtedly getting smarter. But they’re at their smartest when they work as a team. Very soon, researchers will be able to study their every move in the very latest lab to be built at TU Delft: the Cyberzoo.
“To survive in this zoo, robots have to make their own decisions, such as how to go about finding new energy.”
So just imagine: you’ve sneaked into the Aeroplane Hall of the Faculty of Aerospace Engineering (AE) at night with a cunning plan to show your friend your latest conquest, the F16. A bit like ‘Come and see my new moped’, but even better. As you creep furtively through the dark hall trying to stifle your giggles, you hear a noise ahead of you. You go to investigate, like two veritable Sherlocks. Behind a black gauze net, you see a swarm of six-legged beasts rambling around the floor, while a dozen giant dragonflies flutter happily above them. They aren’t at all threatening and simply move aside when you pass through their terrain. You gaze at each other in pure amazement. It’s like a zoo; a zoo inhabited by robots.
Tiny birdbrains
This sums up the dream of Chris Verhoeven, one of the three theme leaders in the TU Delft Robotics Institute, who is in charge of the theme ‘swarm robots’: “We used to think that intelligence could be best accomplished in one large robot. But we’re now moving towards the idea that a group of smaller, simpler robots is actually much smarter.” Verhoeven’s ideas are inspired by nature: bees, ants or fish working as a group are able to carry out much more complex tasks than when they work as individuals. A swarm of bees, for example, can locate all the flowers in an area covering 500 square kilometres. “Our swarm robots must also be able to perform at this level,” says Verhoeven, associate professor in the Faculty of Electrical Engineering, Mathematics and Computer Science (EEMCS), and part-time member of AE staff.
Verhoeven outlines his ideal vision. “We make sure that the robots stay in a group, without getting too close to each other, as if you’d joined them together with invisible elastic bands. If you send a swarm of these robots into a collapsed building, they will sift through every nook and cranny, constantly updating each other on their findings. A robot placed at the entrance will inform rescue workers if they find a victim. The robots can then shine lights along a path, leading the rescue workers to the exact spot,” he explains using a concrete example. “It doesn’t matter which robot finds the victim, as long as they operate as a group to complete the task in hand.” This is, however, still very much in the future.
Forest fires
Studies of how robots of the same type can work together have already been carried out. Last year, TU Delft and the fire brigade tested twelve X100s. These are flying robots, which can work together to pinpoint the site of a forest fire, thereby enabling the fire brigade to step in and extinguish it immediately. A team of fire-extinguishing robots or a robot rescue team would be even more useful to the fire brigade, but researchers have yet to discover whether these different types of robots can work together effectively.
Things are about to change. The contours of a new lab are already materialising in the Aeroplane Hall, a lab in which the beasts can ramble around the floor or dart through the air. An area measuring ten-by-ten metres has been set aside for a steel construction in which a black, seven-metre wide-mesh net will be hung. This is the distance from floor to ceiling. The net is there to confine the flying robots and stop them escaping, rather like an aviary. A Mars-type landscape will be created on the floor using cat grit, pebbles and larger stones. “At the moment, you can still see the contours of the lab, but it will eventually become a real Cyberzoo,” says Verhoeven.
Twelve high-tech cameras have been mounted high above the lab. Another twelve will soon be added, making 24. Each of these cameras is able to shine a tiny infrared lamp onto a robot. Or to be more precise, onto a small white patch that has been attached to the robot, reflecting light in the direction of the camera so that each robot can be located and tracked. This will enable the very first 3D simulations of robot interaction, which can then be analysed to see how the different robots move and work together. But for various reasons, this is turning out to be a highly complicated business.
Radio pheromones
One of the most difficult aspects is working out exactly what each robot does. When people work as a team, there is usually a leader who gives instructions telling the others what to do. This is not the case with ants, although these insects can carry out highly complex tasks together. How does this work? “Ants mark the spot where they find food with an odour, a pheromone. If a couple of ants chance upon a shorter route and more ants follow them, the odour will become stronger and attract even more ants,” explains Guido de Croon, assistant professor in the micro air vehicle lab at AE, one of the labs in the Delft Robotics Institute.
Verhoeven: “The robots will deploy the same sort of interaction, but using radio signals instead of real odours. We’ve dubbed them radio pheromones.”
But the robots will also work with real smells. Some robots are to be fitted with an E-nose, an electronic nose that will allow them to ‘smell’ acetylene gas, for example, a gas emitted by fruit when it is ripe. This technology would enable a DelFly Explorer, the world’s most autonomous micro airplane, to locate ripe pears, making it a sort of fruit fly that will soon be in great demand. Verhoeven reflects for a moment, remembering King Willem-Alexander´s first trade mission, when he himself travelled to Germany with the royal delegation. “Someone from TNT was trying to imagine a whole swarm of DelFlies sniffing the parcels before they are loaded onto the aeroplane. After all, money, drugs and explosives can all be traced by smell.”
Despite only weighing 20 grams, this type of flying robot´s ability to avoid obstacles allows it to explore unknown spaces. The DelFly micro is even lighter, weighing in at just 3 grammes. “It weighs the same as a sugar lump,” says De Croon. The lightness of the robot is proving his greatest scientific challenge. “Providing small robots with intelligence requires a different strategy than for larger robots. It is much trickier to make small robots fly autonomously, as they carry less computing power, fewer sensors and a poorer camera. So we need to develop a highly efficient type of artificial intelligence that will enable them to perform additional tasks,” he continues.
Energy efficient
A zebro (a six-legged autonomous robot) lurches around the middle of the pen. Two lab coats are lying on a rickety wooden platform that has been placed in the pit containing cat grit and stones. They are supposed to represent the torn airbag of a lander, which has just landed safely on Mars and from which the zebro has emerged. The zebro is trying to clamber over the airbag and reach the improvised Mars landscape without getting its rotating legs stuck in the grit.
And as if this wasn’t difficult enough, a DelFly is hovering above the zebro, trying to land on top of its larger counterpart. Although Verhoeven is not quite sure why this may be useful, he hopes that the DelFly will eventually be able to draw energy from the zebro like a mosquito drawing blood. A zebro can transport more batteries and therefore more energy, so a swarm of zebros would be able to keep a swarm of DelFlies airborne. This in turn would dispense with the need for the DelFly to recharge its batteries at a special charging point, enabling these tiny robots to cover greater distances.
This form of future collaboration is a perfect example of the ecosystem that this visionary theme leader has in mind. “The robots must decide for themselves how to obtain the energy they need,” he says, outlining a scenario in which robots extract energy from other robots like a pack of carnivores, while their flying counterparts do the same like a swarm of mosquitoes. But he also envisages herbivores topping up their energy levels at charging points. His favourite image involves collaboration with nanosatellites (such as Delfi-C3 and Delfi-n3Xt) for long-distance communication and global imaging.
Emergency help
“When there’s an emergency in a remote area where people need food, water and medication, satellites analyse the affected area and pinpoint the spots where help is needed most. The flying robots can pass on the details of these areas, locate the victims or pinpoint the exact site of a blaze, so that the walking robots know exactly what type of help is needed and where,” says Verhoeven.
Although the team leader sees endless opportunities for deploying robots in everyday life, he would prefer to see other people coming up with practical ideas for applications. “Instead of holding lengthy discussions about setting up this lab, we just started building it. If you start from a theory, you never actually come up against certain details. Starting with the practical side, like we did, can be incredibly inspiring. This lab is the practical and concrete proof of our technical prowess. If someone like Mark Zuckerberg (the founder of Facebook, ed.) were to drop in, there’s a good chance that he would suggest the ultimate killer application. That’s how technology works.”
We are actually closer than we think to seeing robots moving around at random locations in the middle of the night. In Verhoeven’s words: “At some point, we’ll stop noticing robots roving around and flying above our heads. In much the same way as we’ve stopped noticing smoke detectors on the ceiling, we won’t bat an eyelid when we see our mini-guardian angels doing their bit to make our lives easier and safer.” He is now looking forward to the festive opening of the Cyberzoo in March.
‘Cyberzoo’ on www.youtube.com
Comments are closed.