Smart Sensors are Modernizing Connectivity and Analytics
The internet of things (IoT) is getting smarter due to increased processing capabilities of embedded processors, distributed processing and data reduction capabilities of embedded systems and gateways. Smart buildings, smart cities, and now even wearable devices can use smarter sensor processing to compile and process data locally, allowing more efficient utilization of valuable communication bandwidth, as well as allowing intelligent decision making locally versus having to go to the cloud and back.
The capabilities of embedded devices are increasing due to the increasing capability of modern processor architectures, faster clock speeds and a larger memory footprint. Many embedded processors now include ARM processors that allow for single-cycle DSP operations and very capable libraries such as the CSMIS DPS and Neural Net libraries.
Smart sensors have traditionally been functionally simple devices converting physical variables into electrical signals and transmitting the results with little analysis to the cloud.
These sensors are multi-functional devices which possess the following properties to perform as an IoT component:
Low cost: as to be deployed in large numbers
Physically small: as “to blend in” with the environment
Wireless: as a wired connection would not likely be possible
Capable of self-identification and self-validation
Low power: as to be able to function for years without a battery charge
Strong: as to require little or no maintenance
Self-calibrating or capable of receiving wireless calibration
Capable of data pre-processing and possibly data reduction
Digital signal processing and decision-making capabilities using DSP and local AI algorithms
Information from multiple sensors can be combined or “fused” to reach conclusions such as the onset of mechanical failure. In some cases, the multi-sensor functions are available in one device and processed on that device or processed by the local gateway.
Consider a physical process that generates large amounts of data that periodically needs to be analyzed, but the power profile and the physical communication protocol selected do not allow for larger amounts of data to be transmitted. A powerful processor can perform filtering, spectral analysis, convolutional data reduction, etc. to reduce the amount of data to be transmitted without utilizing excessive power or bandwidth.
The signal processing architecture of the system can now include data reduction on the sensor. For example, in the case of a convolutional neural network, the initial convolutional processing in the fits N layers occurs on the sensor, reducing the data that needs to get transmitted to the cloud for continued deeper layer processing and final decision output. In the same manner, lossless or lossy compression can occur on the sensor or gateway, reducing data usage and cloud processing overhead.
If desired, the sensor could function where it only transmits data if the measured variable value changes significantly from previous sample values, via the output of sophisticated signal processing on the sensor gating non-important data from utilizing precious power and bandwidth.
Smarter IoT sensors can be used to provide greater visibility into existing processes and workflows, identifying items, locating them, and determining their environmental conditions.
Smarter IoT sensors can also be embedded into larger products to improve processes or the products themselves and enable wider and more effective data gathering and analysis.
