A robot can now use the unique signal emitted by a RFID tag affixed to a particular object in order to gradually navigate its way towards the object and locate it, without using any form of visual identification. The days of the affordable personal robot assistant in the home appear to have come a step nearer.R
US robotics researchers have recently discovered that a cheap and simple way of getting a robot to orientate itself in its environment and track down specific objects is to use Ultra-High Frequency Radio Frequency Identification (UHF RFID) tags. Any item can be tagged for UHF RFID at very low cost and the robot can then navigate its way towards the point where the signal is strongest in order to locate the item in question. This method is a completely different approach to the 3D-visualisation tracking systems developed to date, which require advanced algorithms and massive computing power. Those traditional – and much more complex – robot orientation methods basically require the robot to model its near space as a basis for orientation and navigation. The new RFID approach has been pioneered by robotics scientists Charlie Kemp, a Georgia Tech academic researcher, Travis Deyle (formerly of Georgia Tech and now working for Google) and Matthew Reynolds of the University of Washington, using a PR2 robot from Menlo Park research lab and incubator Willow Garage. The robot’s key components are two mobile directional antennas, which can receive RFID signals from quite far away, enabling the robots to track down objects across average household spaces.
The ‘hot or cold’ approach to hunting objects
RFID technology provides a method of precise identification for any tagged object or person. If you place a RFID tag on a wall or item in the house, the PR robot can detect the unique signal. To find the object efficiently, however, the robot needs to be a bit smarter than that. In fact the research team are currently working on an algorithm to refine the way the robot identifies the RFID tag and navigates towards it. The basis of the system is the received signal strength indicator (RSSI), which varies enormously according to the object’s ‘pose’ (position + orientation), material properties and surroundings. Using its shoulder antennas, the robot will turn in the direction that maximizes the received signal strength indicator associated with the target tag, not perfectly plotting a course in advance but zeroing in on the target object by a process of trial and error – rather like the traditional childhood hide-and-seek game of ‘Hotter/Colder’, say the researchers – with the tag emitting a stronger signal as the PR2 gets closer to the target.
The advantages of UHF RFID over visual technology
The first practical application of this research will probably be for robots tasked to assist dependent people, e.g. fetching medicine from a cabinet for a patient who needs it. In addition to their capacity for unique identification – there are virtually no ‘false positives’ and therefore a negligible error rate – and very low cost, another advantage of RFID tags is that they are very small and can therefore be affixed to small items. This means this approach is particularly well-suited to robots tasked to find and deliver objects. The research team is now working on a new search algorithm that will improve a robot’s ability to locate and navigate towards tagged objects. The drawback with the visual approach for robot search previously developed, based on cameras, lasers and 3D-plotting, is that lots of objects such as medicine bottles look very similar but must be uniquely identified, requiring a very complex algorithm. The RFID tag method would appear to involve much less risk than a visually-based search that the robot will confuse items and perhaps deliver the wrong bottle.