American scientists have replaced the optical perception of robots by scanning RFID tags. This development could also facilitate the communication of drone swarms.
A test run at the MIT: The bottle and lid are provided with RFID tags and can thus be located by the robots – with every movement.
In some situations a robot has a similar problem as humans. If, for example, he has to recognize objects for complex tasks, his visual perception – machine vision – is often insufficient. For a large number of objects that may obscure each other, it is difficult for robots to find sufficient visual starting points. Scientists at the Massachusetts Institute of Technology (MIT) have now developed a solution for this. They used radio frequency signals to identify objects.
The advantage of Radio Frequency Identification (RFID) is its independence from visual perception. Robots can use RFID to identify individual targets on a large scale, even through walls. The researchers use a system called TurboTrack. In this case, a reading device initially transmits a wireless signal which is reflected by the RFID tag and other objects in the vicinity and is received again by the reading device. An algorithm scans all reflected signals to filter out the RFID tag’s response – each day reflects the signal in an individual pattern. The permanent emission of the signal causes the robot to understand the movement of the object. During test runs, the robots found the marked objects on average within 7.5 milliseconds. In terms of localization, the accuracy deviated by a maximum of one centimeter.
Imaging technology as a model for the calculation of RFID signals
The idea of using RFID tags for the localization of objects is not new. In previous experiments, however, the scientists had to cut corners because the results were too inaccurate or the search took too long. For this reason, the team at MIT has been inspired by an imaging technique for their system: for high-resolution images, individual images created from different perspectives are combined into a single image. The researchers have applied this principle to radio signals. When something moves, you can track it, multiply its position, and match the results so that the motion contributes to more accurate localization.
All signals are transmitted at the speed of light. So the system can measure the time the signal takes to get back to the transmitter and calculate the distance. As the RFID tag moves, the signal angle also changes slightly. The algorithm can also use this to understand the movement of the object. In addition, it includes all other signals in its calculations that are reflected by other objects in the environment, including their changes. By constantly comparing the distance measurements of all the signals, the tag can be found in a three-dimensional space. All this happens in a fraction of a second.
Better communication for drone swarms
The most accurate localization of objects should enable robots to take on complex tasks. In manufacturing, for example, they could pick up, assemble and finally pack products along an assembly line. The scientists have therefore reviewed their development in an application scenario: they attached one RFID tag to one bottle and a second one to the corresponding lid. Then they had a robot – which localized the object without machine vision – holding the bottle, a second one locked it with the lid. This worked just as fast as traditional experiments in which the robots controlled their actions through machine vision.
Another potential application is drone technology. For example, for search and rescue tasks drone swarms are used to cover larger areas. This brings many challenges. Among other things, the drones must be able to communicate with each other and coordinate their movements exactly. This is often hampered by rough terrain and obstacles such as walls and trees. The new RFID system could help drones to better recognize and position the each other places can. In a test run, the researchers were able to successfully track RFID-tagged drones, how they flew and maneuvered.
Further developments in the field of robotics:
- Industrial robots play Jenga and learn in real time
- Electronic skin improves robot control
- Machines learn moral judgment