Modern advances in computing hardware and machine learning algorithms have enabled many new applications in robotics, human-machine interfaces, medical devices, and much more. But to be of much use, these systems need information about certain aspects of the world around them, which is provided by some type of sensor in most cases. As a general rule, the better the resolution that these sensors can provide, the better the results that can be produced by the algorithms. As the old saying goes: garbage in, garbage out.
For a great many use cases, the most effective way to collect data is via a sensor matrix. These dense grids of sensing elements have the ability to provide high-resolution and multidimensional measurements to downstream processes, which allows for better accuracy and precision. Unfortunately, the design methods employed in building these sensor matrices leave much to be desired. For starters, wiring them quickly becomes problematic as an M x N sensor matrix requires M x N wires. Each of these wires is a potential point of failure, leaving these devices prone to malfunctioning. Moreover, this configuration wastes a lot of space, which can make such designs unsuitable for applications where device size is a critical factor. Further issues arise due to the fact that measurements from such systems are captured using time-division multiplexing, which reduces the temporal resolution of readings.
To address these issues, developers have created matrices with redundant connections and components, involving sensor nodes configured in mesh networks. But this approach also has its issues in that it requires substantial computing power and greatly impedes miniaturization. A completely different approach taken by researchers at the Hong Kong University of Science and Technology may help to alleviate these problems in the future, however. They have developed a sensor matrix that utilizes only a single wire to carry signals from all sensors.
The team took inspiration from tonotopy, which is the process by which different sound frequencies are transmitted from the ear to the auditory cortex in the brain for processing. In the same way that each hair cell in the cochlea is tuned to a different frequency of sound, each sensor in the matrix is assigned a unique sine wave frequency, which serves as its identifier. The measurement captured by each sensor is then encoded by modulating the amplitude of the signal. As this signal travels through the single wire, measurements from each sensor are made available to downstream processing units, with the help of a Fast Fourier Transform algorithm, by identifying each unit’s unique frequency and decoding the information encoded in that signal’s amplitude.
In a demonstration of their technology, the researchers built a glove with a matrix of both pressure and temperature sensors. By grasping a cup containing liquids of varying temperatures, they showed how the device can accurately track both pressure and temperature measurements across the entire surface of the glove using only a single wire for data. A similar experiment demonstrated how an array of strain gauges can be placed on the wing of an aircraft to monitor for unacceptable levels of strain that could cause a loss of flight control.
While this innovative sensing matrix may solve a number of existing problems, there is still more work to be done. First, the bandwidth of the system could be a problem for some applications, as the sensors must all be sampled in sequence, not simultaneously. Furthermore, the reliance on off-the-shelf sensors leads to limitations in the device’s possible resolution levels, because these individual sensing elements are only just so small. In any case, this is a significant step forward, and with further development could have important implications in a wide range of fields.