AI and traffic management smart or naïve? Artificial Intelligence (AI) is becoming increasingly successful. To the public, AI sometimes appears as a miracle cure for all difficult problems, including the great challenges that traffic management faces. But is this realistic? Whatever else we know, we know one thing: AI is going to play some kind of role in urban mobility. AI systems for traffic management scenarios There are many kinds of AI. The successes we are currently seeing are occurring with Large Language Models (LLMs) such as ChatGPT and DeepSeek, and with Vision AI for image recognition systems. LLMs and Vision AI work with neural networks. The power of neural networks is that they can learn to make connections, either independently or under human guidance (supervised learning). In the future, neural networks should be able to automatically generate traffic management scenarios for a platform like MobiMaestro. An AI module could, for example, independently adapt the available physical and digital traffic management instruments to current circumstances, guiding and influencing the traffic. But this is not a reality yet. Large quantities of training data are required to train neural networks, and there is not enough available data yet in the domain of traffic management. But the fact that this can’t be done yet at the moment does not mean that it is impossible. Technolution Move is focusing in projects and research on collecting useful data for effective training methods. Vision AI for traffic applications Using Vision AI in traffic-related applications has been successful for some time now. Thus, we are seeing Vision AI in autonomous driving. Vision AI creates an image of the surroundings of the vehicle to keep it within lane markings and away from other vehicles and road users. This technology is now quite mature. In China and the United States, cars equipped with computer vision are even allowed on the road fully autonomously. At Technolution Move, we are using Vision AI in our FlowCube traffic sensor and in customized solutions for our clients. For instance, a specialized camera and Vision AI measure the draft of passing ships at the IJmuiden sea lock. The system reads draft markings on the vessels and compensates automatically for swell and ship movements. If the draft exceeds the limits for passage towards Amsterdam through the North Sea Canal, the system issues a warning. 21
22 Online Touch Home