Embedded technologies are indispensable for coping with socio-political challenges such as climate protection, demographic development or increasing resource efficiency. They also form the basis for the majority of innovations in important industrial sectors.
Computers are disappearing from our perception and are at the same time omnipresent. Advances in microelectronics mean that intelligent functions can now be implemented decentrally, wherever they are needed. Such “embedded systems” generally operate with a high degree of autonomy. This concerns the detection (sensors) of the environment, the interaction (actuators) with it and the communication with other embedded systems or computers. The required functionalities are increasingly provided by artificial intelligence (AI) and pattern recognition algorithms.
The data processing takes place in a central data center (cloud) or, increasingly, at the point of origin itself (edge computing). The latter is particularly suitable for time- and safety-critical applications, but also helps to reduce the immense amounts of data to be transferred. In a largely automated, adaptive production environment or in autonomous vehicles, the direct processing of data in the device even proves to be essential, for example to avoid impending machine failures or accidents. A prerequisite for this, however, is “intelligence” that is specially tailored to scarce resources.
For example, AIfES (Artificial Intelligence for Embedded Systems) from the Fraunhofer IMS enables the development of small, intelligent microelectronics and sensors that do not require a connection to a cloud and are capable of learning on their own. AIfES contains a fully configurable artificial neural network (KNN) with feedforward structure, which makes even deep neural networks possible.
Since with AIfES the processing takes place “offline” on the device, no sensitive data has to be transmitted. Privacy is therefore ensured. But to protect the networked and increasingly complex systems from hackers, preventive and seamlessly integrated security solutions are needed. And this applies not only to companies, but also to politics. Just think of attacks on critical infrastructure, for example in the energy sector, or connected-car applications. However, classic cyber-security approaches such as anti-virus software or patches and updates do not work for IoT devices. Security requirements for embedded systems must therefore be taken into account during the development phase (security by design).
In the future, the transmission is to be much more secure via 5G. However, the networking of machines and devices with the new mobile communications standard will gain momentum mainly because of its ultra-short latency times and high data transmission rates. In addition, there will be much lower energy consumption.
The analyst firm Gartner expects the 3.5 million 5G IoT nodes to increase to 49 million this year by 2023. From 2023, embedded systems in connected cars are expected to account for the lion's share. Until then, outdoor surveillance cameras will be the main drivers of growth. However, 5G will also be used in sensor cloud systems in the future to monitor highly dynamic production processes more thoroughly and control them adaptively.