ROBOTS will never be much good at household tasks such as pouring coffee or polishing shoes unless they can know their position accurately. But unlike robots that operate outdoors, domestic droids can鈥檛 pick up the weak satellite signals from the Global Positioning System, so designers who want to put their creations to work have almost literally hit a brick wall.
But that鈥檚 all set to change, courtesy of a new 鈥渋ndoor GPS鈥� system from the Massachusetts Institute of Technology. One of its inventors, Seth Teller, believes it could dramatically improve robots鈥� performance in homes, shops and offices.
GPS receivers work out their position by comparing the arrival times of radio signals from four or more satellites flying overhead. Teller and his colleague Hari Balakrishnan have made an indoor equivalent that can easily be set up in any room.
Advertisement
The system is based on a network of beacons that broadcast signals into the room. 鈥淟istener鈥� devices on the robot or any other device that needs to know where it is then use the signals to determine their position, which they can also broadcast if need be.
The beacons broadcast a radio signal followed by an ultrasound chirp. The radio signal tells a robot which beacon is broadcasting, and warns it to start listening for the slower, ultrasound signal. The robot can work out its distance to the beacon by comparing the arrival time of the radio and ultrasonic signals.
By repeating this process for a handful of beacons, the robot works out its position and orientation. 鈥淚t鈥檚 accurate to within a few centimetres and 2 or 3 degrees of arc,鈥� says Balakrishnan. Because the beacons chirp, the team have dubbed the system a 鈥渃ricket鈥� network.
The project is linked to MIT鈥檚 Project Oxygen, which aims to make computing ubiquitous 鈥� essentially by building computers (and crickets) into everyday objects so that people are not even aware of them. They liken it to the way electric motors, which are now built into so many devices such as dishwashers and food processors, were sold separately a century ago.
Teller thinks crickets will go a long way towards solving one of the big problems in robotics. A robot attempting to grasp an object faces many difficulties: it must calculate a trajectory for its arm to pick up the object without knocking over anything else. It must then grab the object firmly but not so hard that it crushes it. 鈥淭he reason you never see robots clearing tables is that it鈥檚 very difficult to do,鈥� says Teller.
But a cricket network could make the robot鈥檚 task much easier. A coffee mug, for example, fitted with a cricket listening device could broadcast its identity, position and orientation. The robot can then look up the mug鈥檚 properties such as its shape and strength from a pre-stored memory 鈥� or from the Web, say. Calculating an appropriate trajectory for its arm and grasping the mug then becomes straightforward for the robot.
Powering the cricket devices will be a big challenge for the MIT team, though, predicts Jamey Hicks, a computer scientist at Hewlett-Packard鈥檚 Cambridge research lab in Massachusetts. He says the radio and ultrasound devices are 鈥渢oo big and power-hungry鈥� to use in large numbers right now.
