The Internet of Things (IoT) is growing rapidly all around the world. New devices are continually being added, all collecting a variety of data and transmitting them (often wirelessly) to edge devices, which in turn relay the data to the cloud for further processing. It is estimated that in a few years IoT devices will be responsible for over 20% of global energy consumption. Minimizing the energy use of IoT devices must therefore be recognized as a clear objective for climate protection.

The energy consumption of each individual IoT node is also a limiting factor on the useful life of the device. When devices are used in hard-to-reach locations and cannot be wired up, minimizing energy consumption is essential to prolonging the service intervals.

Energy use must therefore be considered in the early development of a new device. Important architecture decisions are made already during the concept phase, which must take power consumption into account as a design parameter. For example, when defining the energy supply concept, designers must choose whether the energy will come from a rechargeable battery, a long-life disposable battery or energy harvesting. It is necessary to consider both the static, integral energy demand, as well as the dynamic demand based on the specific usage scenario.

A number of other conceptual questions must also be answered at an early stage since they have an impact on the required power supply. For instance, a balance must be found between using power-hungry transmission modules for uploading extensive raw data and energy-intensive compression of data prior to sending. Local aggregation of sensor data and extraction of interesting features is another way to reduce the volume of transmitted data, although at the cost of algorithmic complexity. Sophisticated machine learning or AI algorithms can hardly be implemented on simple microcontrollers. If these functions are shifted to the edge nodes, more data must be transmitted to those devices.

Questions about the operating regime also come into play here: How often must sensor data be collected to achieve a given level of precision? This factor influences decisions about communication from the sensor device to the edge node. Sending data more frequently increases the reliability of the transmission, while less frequent transmission must be secured with error correction measures. In the worst case, sensitive data must be additionally encrypted before transmission.

Which transmission protocol to use is an important question in networks consisting of many IoT nodes. Synchronous protocols only exchange information at specifically defined times, and the nodes enter sleep mode during the intervening intervals. With asynchronous protocols, the nodes only wake up from sleep mode when data is actually available, but they must be continuously ready for such activation. Depending on the required data rates and operating ranges, a variety of protocols are available: from Zigbee and Bluetooth Low Energy (BLE) to special IoT protocols such as Long Range (LoRa), Narrow Band IoT (NB-IoT) or even 5G variants for massive Machine Type Communication (mMTC).

It is therefore clear that the energy consumption of IoT devices is a factor in many design decisions and that the operating mode requirements must also be taken into account. This makes it necessary to compare various alternatives during the concept phase with the help of appropriate tools. The energy consumption of individual components in a given operating mode must be considered, and an energy-sensitive data flow model is needed at the architecture level. The required energy consumption naturally depends heavily on the hardware used. General transaction-based processor models must be supplemented with information on the energy consumption of communication and computing steps. The design space can then be searched by simulating various scenarios from the perspective of energy demand.

The vast range of potential IoT components and algorithms makes the complexity of this design task enormous. Increasing automation is therefore required to find an optimal solution between the competing goals of performance and energy conservation. Such automation is only possible with an end-to-end digital description of the system, which must be available from the concept phase through to implementation in order to analyze and optimize the interactions between components. Relevant approaches can be found in the methods of contract-based design, which describes the system components as agents, placing demands on each other and guaranteeing specific services. These descriptions allow step-by-step refinement during the design process as well as automation of the design of complex systems by means of optimization methods.

In general, methods of model-based systems engineering (MBSE) provide a solid basis for early decisions in the design of complex hardware/software systems. The associated models are developed at a high level of abstraction and can be re-used later (enriched with implementation details) for verification of the complete system.

Considering the unbelievable number of IoT systems that are forecast, tremendous energy savings can be expected. However, an intelligent MBSE approach will also improve the useful life of the individual systems.