Adapt dynamic systems in real time

Knowledge-based logic systems for traffic control systems, production sites, train logistics and more

Knowledge-based logic systems help find optimized configurations for switchgear and railway interlockings. Until now, however, they had never been used to control highly dynamic areas like road traffic or communication networks. The research project DynaCon, which was completed at the end of 2020 and included Siemens as a participant, has now found solutions that are not only feasible but are also faster and more reliable than other AI methods.

The road traffic of tomorrow is networked. Traffic lights, message signs, buses, trains, and eventually even self-driving electric cars will all be busily communicating in order to identify the fastest route from A to B, ensure a minimum safety distance, avoid traffic jams, and flexibly respond to accidents and road work. However, this requires that cameras in cars and at intersections, electromagnetic sensors in the road surface, and transmitting units in vehicles transmit a continuous stream of data – data that must be constantly evaluated and applied to the traffic situation.

Traffic management in real time will be achievable over the next few years once traffic control centers start acting more autonomously. This requires intelligence – artificial intelligence – which comes in several forms. In the case of machine learning, for example, the more data that’s fed into the system, the better it can recognize patterns and regularities. Simulation models that enable predictions based on empirical data are another form of modeling reality.

Reconfiguration in real time

But in cases where complex systems need to reach an optimal level of performance – such as a smooth flow of traffic – as quickly as possible, knowledge-based logic systems are the fastest way to arrive at a solution. They apply area-specific data and rules in order to instantly find workable solutions. For over twenty years, industry has been using this approach for non-time-critical applications such as identifying configurations that reduce the manufacturing costs for new telephone systems or railway interlockings. A similar situation is the conversion of production plants that require constant reconfigurations over the course of their deployment.

What has been missing until now is the option for flexibly controlling dynamic systems in real time with the aid of knowledge-based logic systems. This was the focus of the DynaCon (dynamic knowledge-based (re)configuration of cyber-physical systems) research project, a three-year project completed at the end of 2020 that was led by the University of Klagenfurt and included five other partners from industry and academic research, one of which was Siemens AG.

Reduced development and maintenance costs

The goal of the project was to use data streams to control highly dynamic systems like traffic, communication networks, and railway logistics without an appreciable time lag. Because all these systems have networked mechanical and electronic components that can be controlled digitally, they’re also referred to as cyber-physical systems. “They require constant reconfiguration in order to optimize their overall behavior,” says Gerhard Friedrich of the Institute for Artificial Intelligence and Cybersecurity at the University of Klagenfurt.

So how does this type of knowledge-based logic control system work? The DynaCon team’s first step was to prepare the knowledge about a specific plant or system by formulating it as logical statements stored in a database. These statements range from the simple – for example, when a traffic light at an intersection is green, the traffic light for the cross street must be red – to formulas for traffic density that change the timing of traffic lights when a certain threshold value is reached. 

Detailed system knowledge is necessary

Each individual step is important for finding the solution, because each decision eliminates other available options. For example, the choice of a specific car body on an assembly line limits the range of additional parts that can be installed, whether it’s a shock absorber, radiator grill, or windshield. “Achieving this type of control requires detailed system knowledge,” says Andreas Falkner from Siemens’ Technology research department. “At the same time, the logic rules that are used can generally be applied to other areas as well, with only a change in variables.”

The algorithms and programs developed within the framework of DynaCon have already been used multiple times in test runs, including for controlling an Internet service provider’s communication networks, for freight traffic logistics, for power distribution, and for urban traffic. The latest development in control technology is not only much faster than earlier solutions, but also results in higher-quality solutions and reduces development and maintenance costs. Of course, these lightning-fast adaptations wouldn’t be necessary if everything proceeded according to preconceived plans, day after day, but such is rarely the case.

Dynacon uses smart recycling

How did the DynaCon team manage to design knowledge-based programs in such a way that they can control dynamic systems in real time based on numerous data streams? More powerful hardware is not enough. It also takes efficiency. This process of instantly identifying solutions uses only the parts of the knowledge base that it actually needs. If what is currently being assembled on a production line is a sports car, it has no interest in data relevant for delivery trucks. It is also constantly filtering out any relevant environmental conditions. For example, pavement that’s wet with rain is relevant whereas a single cloud that momentarily obscures the sun is not. DynaCon also practices smart recycling: When a new, situation-specific solution is found, it is generalized and incorporated in the knowledge database for use when similar situations arise.

Can be easily integrated into various control systems

Now that the research project has ended, the DynaCon team can not only point to patent applications and invention disclosures. The innovative control programs are now set to be used in new pilot projects – and no significant additional effort is needed. According to Falkner, “The program can be imported into a variety of control systems without any difficulties.”

Christian Lettner, Hubertus Breuer, June 2021

Subscribe to our Newsletter

Stay up to date at all times: everything you need to know about electrification, automation, and digitalization.