One of the key ideas behind the IoT is the use of sensor networks to gather data. But when there may be thousands of sensors in a network, each sensor needs to be as low cost as possible.
However, low cost sensors often have limited energy and computing capacities, so making the most of these limited resources is important and Austrian researchers say one way in which this can be done is through the use of time synchronisation.
The approach learns the behaviour of sensor clocks and this is said to makie it efficient in terms of energy and computational resources by minimising the amount of time the sensor needs to use its radio.
Between them, Wasif Masood, Dr Jorge Schmidt and Professor Christian Bettstetter, from the Institute for Networked and Embedded Systems at the Alpen-Adria-Universität Klagenfurt, have developed a technique that reduces the additional effort of synchronisation between the oscillators of individual sensors.
Dr Schmidt, pictured, explained: “With a group of friends, we already know who is usually late. Therefore, the coordinator of such a meeting could tell individual friends different times in order to intercept the delay. This is exactly what the newly developed technique does – by using time series analysis, it learns the behaviour of the sensor clocks and can anticipate or correct future deferrals before asynchronicities begin to develop.
“While the idea of learning behaviours to predict future corrections is not new, we have shown that behaviour models extracted from our time series analysis work very well with commonly employed wireless sensor devices.”
Author
Graham Pitcher
Source: www.newelectronics.co.uk