System Architecture Layers

The Internet of Things connects the physical world to data servers in the digital world and presents a value to the customer. A simple model to illustrate the logic is presented below:

The physical Thing can reside anywhere as a fixed device – e.g. factory machinery – or a moving object – e.g. refrigerated containers. Sensors register a physical property on the physical thing such as temperature or mechanical position of a piece of moveable machinery. Actuators are operating the Thing, acting on digital control commands to open, turn, cool, move etc. the Thing. Often layer 1 & 2 are integrated and run an autonomous closed loop control system to steer for instance servo mechanisms or maintaining the set point temperature by operating a hot water valve.
Data collected by sensors and control commands from the Digital Service are sent and received through the connectivity layer, which will be selected to cover the physical distance of the transmissions and optimal bandwidth needed for the purpose. Connectivity can be divided into e.g. local wireless connectivity on the site then using public internet for connecting to the servers in the digital world.
The installation in the physical world is also referred to as The Edge.
In the digital world, The Cloud, data are stored and normalised to enhance digital raw formats with informative metadata and supplementary information such as time stamp, location tag, sample rate, etc. hosted on a central computer system. “Cloud” is just another term for very scalable Server farm but the word indicates an abstraction of minor concern to where processing is conducted as long as the end user is experiencing the desired output on a dedicated PC, tablet, or mobile terminal.
Analytics models data using common statistic tools such as trend lines, mean, variance further enhanced by applying correlation and regression algorithms – all with the purpose of learning patterns and later compare those across data sets. Different training methods for prediction are applied until the model achieves a desired level of accuracy on the training data.
The Digital Service presents data as direct readouts – timelines, gauges, heat maps, etc. – as well as visualising those patterns learnt during analytics. This can be presented as large display dashboards or sent to a handheld device for direct customer or operator interaction.
A factory quality operator can now from a dashboard watch a colour map of components indicating the risk it soon fails based on their manufacturing history compared to years of experience the machinery supplier has collected from similar parts across a world-wide install base.
The paragraphs below exemplify further or elaborate on certain characteristics to consider for each layer:

Physical things

The device or apparatus having a physical presence: Electrical conveyer belt, hydraulic position servo, automatic door, drainage pump, blood pressure meter. It is the entity which must be measured or controlled.


Sensors send out an electrical signal as a reading of Position, Temperature, GPS module, Power Dissipation, Pressure, Vibration, Sound, etc.
Actuators consume energy to move or activate Air Valves, Hydraulics, Electrical Motor shafts, Servo Mechanisms, Drainage Pumps, Vibrators, etc.

Energy scarcity

Sensors in remote places may be restricted by very low power available at the measurement point. At any time new ways of minimising maintenance of deployed devices including replacement of batteries are researched, which have led to the advent of advanced energy harvesting principles: a piezo-element generating power from squeezing can catch minimal movements to generate power for a sensor and transmitter; a sensor on a unit with a distinct temperature difference can harvest energy via the Peltier-effect. Yet other devices are striving to collect solar energy for a sealed battery not due for replacement in 10 years’ time.

IoT-relevant components cost erosion

Certain components are prone to a dramatic volume boost in the market as other domains pick up the functionality sparking initially shortage then a fierce competition leading to heavy price erosion. The array of sensors deployed in smartphones such as accelerometers, digital compass, GPS receiver, medium- and high-definition camera, light sensors, microphones, etc. provide a rich source of opportunities for interfacing to industrial or customer equipment for any other purpose.
Once becoming a common feature in the smartphone domain (the 1 billion LTE smartphone shipment mark was surpassed in 2016) a component such as an accelerometer is available at approximately 0.50 USD – in the 1980ies the cost was above 500 USD a piece.

Data collection

Data collection establishes a buffer between the measurements from sensors and the connectivity used for forwarding the data. The bandwidth may not be the same, or other data may be bundled before forwarding. Further the data collection point can perform initial calculations or controls on the data and actuators linked.

Experimentation devices

For experimentation and business verification purposes a suite of general modular devices are available connecting via standard interfaces to sensors and performing data handling/transmitting. This simplifies connecting to external sensors and a choice of wireless connectivity. An intuitive software environment (SDK) supports fast experimentation and remote update deployed to devices on site or in a lab.
A bit geekier experimental approach is to acquire a stand-alone computing board, e.g. C.H.I.P. (9 USD); Raspberry Pi Zero (from 5 USD). As the centrepiece of a vast array of inventions based on Linux, Windows-10 or other COTS Operating Systems the ecosystem is extensive but fragile and meant for continued exploration to gain advantage of already public domain drivers, etc.

Control loop

No matter how reliable and fast the connection would be on a 4G/LTE network, the control of a nuclear power plant would never be controlled via a remote connection – the consequences if disrupted would be fatal.
Very short response times are required for controlling time critical systems as automated closed loops regulating and balancing sub-systems, e.g. timeliness of a factory conveyer belt’s speed to synchronise with other process steps.
Overall control of the system – such as changing the reference temperature set point – is independent of accurate timeliness and can happen from a distance on an operator dashboard or a knob to control the process.
The outer monitoring loop is not directly involved in controlling, rather supervising and reacting to observations over time, e.g. monitoring the mean and variance of temperature stability, comparing across all temperature sensors; benchmarking energy usage with indoor comfort across floors or buildings; and subsequently from best practice suggest corrective actions on the system control.

Connectivity (Digital infrastructure)

Dedicated data transmission protocols have been developed and optimised with the purpose of minimising energy consumption and only transmitting a minimum amount of information in networks where devices can assume sleep mode and wake up when necessary.
Billions of compact radio modules, protocol logic and antennas deployed across handheld devices allow for equipping IoT devices with short (NFC, Bluetooth), medium (WiFi), or long range (3G, LTE) connectivity available as integrated modules at very low cost.
The digital infrastructure should be optimised for the full route from initially connecting a simple sensor towards a hub or router, which in turn is linking the connectivity towards the servers and cloud functionality. Multiple systems may be deployed to complement the requirements of the entire site.

A manufacturer of a standard wireless module highlights: “The ideal for IoT applications and devices, it includes various features such as smart utility metering, smart payment and point-of-sales Systems, and is capable with wearable devices, like action cameras. With end-to-end security features such as secure boot, transport layer, and more, it ready for the LTE World.”

Designing a robust digital infrastructure requires analysis and decisions on 1) selecting modules for optimal data capacity versus cost to 2) market demands and to 3) cover constraints across regional regulation 4) connecting in different geographies 5) across shorter or longer distances, through obstacles, etc.

The above figure represents a hypothetical example of connecting moveable machinery with sensors on a factory floor within short range wireless ZigBee mesh networks. Those in turn are connected via WiFi to both tablets and the cabled (Ethernet) cloud servers via public internet.

Continue reading: go to Cloud and IoT – which to choose?