AI in municipal traffic management: Clear goals determine success

Artificial intelligence (AI) is becoming increasingly important for local authorities. In traffic and mobility management, it opens up new opportunities to reduce congestion, emissions and inefficient traffic flows. However, its benefits are not determined by the technology alone. Without clearly defined municipal goals and measurable criteria, AI remains ineffective. This is the conclusion reached in a recent discussion paper by the Learning Systems Platform. The focus is on a holistic vision of AI-based traffic management for mobility systems based on the needs of local authorities. The paper offers strategic guidance on how AI can be used as a learning management tool to shape mobility in the public interest.

Cities and municipalities are under considerable pressure to act: they are expected to make mobility more efficient, climate-friendly and safer. Traffic flows need to be improved, space redistributed and quality of life enhanced – all while resources are limited and conflicting goals are increasing. AI can help to make smarter use of existing infrastructure and enable data-driven management decisions. To this end, learning systems must be consistently aligned with municipal objectives.

‘AI-based traffic and mobility control only delivers social benefits if it follows clear municipal objectives and is designed to be capable of continuous innovation,’ says Tobias Hesse, German Aerospace Centre and member of Plattform Lernende Systeme, ‘Only when political priorities are translated into measurable indicators can they be controlled by AI systems.’

‘Technically, we rely on model-based reinforcement learning, which allows control strategies to be tested in simulations before they are used in real traffic – essentially a safe laboratory for dynamic traffic systems,’ explains Claus Bahlmann, Siemens Mobility and member of Plattform Lernende Systeme. ‘This avoids learning processes taking place directly in safety-critical, highly dynamic traffic scenarios – you could say we're not performing open-heart surgery.’

Reinforcement learning in particular opens up new possibilities. Unlike traditional, rule-based methods, this method continuously evaluates options for action and dynamically adapts control decisions to changing traffic situations. Practical experience already shows that environmentally sensitive traffic control or the prioritisation of certain types of traffic can be implemented in an adaptive and impact-oriented manner if the objectives are clearly defined.

To enable local authorities to benefit from innovations much more quickly, the paper recommends an open research and development system. This acts as a construction kit. Here, mobility-related data is integrated, new services and control strategies are tested in simulations and then gradually transferred into operation. This not only allows local authorities to retain control, but also avoids dependence on individual providers and accelerates innovation cycles – for example, through the development of new situation awareness services or control algorithms.

About the brochure

The position paper ‘KI-basierte Verkehrs- und Mobilitätssteuerung in kommunalen Mobilitätssystemen’ was written by members of the ‘Mobility and Intelligent Transport Systems’ working group of the Learning Systems Platform. It is available for download free of charge.

A short interview with Tobias Hesse, author of the position paper and member of Plattform Lernende Systeme, is available for editorial use.