Data producing sensors is the new big thing

Without sensors and their tireless digital data production, there can be no AI.

We live a cultural and economic data-driven revolution. Recent progresses on Artificial Intelligence (AI) studies found a profitable application for the large amount of data produced every day by electronic devices. Deep Learning (DL) algorithms, very powerful statistical classification schemes that proved useful in image and speech recognition and are currently applied to many other fields, including customer profiling and marketing analysis, need a large amount of data to perform. Where can we find them?

Digital data are a technological product, they do not exist in nature and are generated through a process called “digitalization”. It starts with a physical quantity, like a magnetic field amplitude, a temperature in a room or the weight of a person. This physical quantity is usually called an analog data, not currently useful for AI applications. For this reason, the analog data is transformed into a digital data. This is done with a device called sensor. A sensor is a device that senses the physical quantity, the analog data (e.g. a temperature), and generates an electrical quantity (typically a voltage or a current). The voltage is fed into another device, called ADC (Analog-to-digital converter) that provides as an output a number, the digital data, proportional to the amplitude of the analog data. A digital data is a finite precision quantity that can be fed into a digital computer and used for AI application purposes.

From this simple description it is clear that in the logic pyramid of AI, a fundamental task is represented by the sensing activity. This generates the digital data to be fed into the DL algorithms that provide as an output the expected predictions for the various use cases (see Figure 1).

Sensing is thus a fundamental aspect of the ongoing digital revolution and it is realized by sensor devices. A simple example of a sensor is represented by the touch screen of our mobile phone. It senses the small pressure of our finger, thanks to a change in the screen electric capacitance (the analog data) that is associated with a change in the electric current. This is sampled and digitalised by an ADC and fed to the phone processor. Another example is represented by the temperature sensing networked devices that monitor the temperature of the car (engine and passenger compartment). There an electric voltage is generated proportional to the temperature, digitalised and sent to the temperature management system of the car.

In recent years we have assisted to a large proliferation of small mobile wireless sensors used to monitor the status of large infrastructures like bridges, buildings, monuments. Structural health monitoring is one of the last frontiers in AI applications. Once again, this is realised thanks to a large amount of vibration data acquired by the distributed network of wireless sensors.

If it is true that data is the new oil, then networked sensors are the new oil rigs. Who own the sensors, own the data. Think about it…