Urban traffic management is a critical issue in modern cities, with congestion leading to inefficiencies that affect everything from daily commuting to emergency response times. As cities grow, traditional methods of traffic forecasting often fall short of meeting the increasing demands for real-time traffic flow management.

However, a breakthrough study on Quantum Neural Networks (QNNs) offers a new path forward, blending cutting-edge quantum computing with deep learning to predict urban traffic patterns more efficiently and accurately than ever before.

The research, titled “Quantum neural networks with data re-uploading for urban traffic time series forecasting”, published in Scientific Reports in June 2025, explores this innovative approach, particularly in the context of Athens, Greece—a city renowned for its traffic congestion.

A New Paradigm for Traffic Forecasting: Quantum Neural Networks

The study, led by Nikolaos Schetakis and colleagues, investigates the application of Quantum Machine Learning (QML) for forecasting urban traffic, marking a significant shift from conventional techniques.

Classical machine learning models, while useful, are often limited in their ability to process and predict highly dynamic, non-linear systems like urban traffic.

Quantum computing, however, offers a potential leap in capability by exploiting the superposition and entanglement properties of quantum states, allowing for more sophisticated data analysis and pattern recognition.

The researchers specifically focus on a novel quantum technique known as data re-uploading. This approach involves encoding classical data into a quantum state multiple times, allowing the model to better capture complex patterns that arise in traffic data.

By repeatedly “re-uploading” data, the quantum model gains a deeper understanding of the underlying dynamics that govern traffic flow, leading to potentially more accurate predictions.

Methodology: Hybrid Quantum-Classical Models

The study employs a hybrid quantum-classical architecture, integrating quantum neural networks with traditional machine learning methods. The key innovation is the use of data re-uploading to enhance the quantum model’s performance. In this setup, quantum circuits are used to process the data, while classical neural networks handle the post-processing and decision-making tasks.

A variety of experimental configurations were tested, including fully connected networks as well as recursive architectures. The researchers compared the performance of these quantum-classical hybrid models against purely classical deep learning models, including state-of-the-art neural networks, to gauge the advantages and disadvantages of each approach.

One of the central findings of the research is that, in traditional fully connected network settings, hybrid quantum-classical models underperformed when compared to their purely classical counterparts. The median scores of the quantum-enhanced models were found to be about 10% worse than the classical models across multiple configurations.

However, in recursive architectures, which are designed to handle sequential data (like traffic time series), the quantum-enhanced models demonstrated superior performance, achieving up to 5% better median scores than their classical counterparts with comparable complexity settings. Additionally, the quantum models required fewer training epochs to converge, indicating a higher training efficiency.

Quantum Models Show Promise in Traffic Forecasting

Despite some early challenges in the fully connected model configurations, the researchers observed that hybrid quantum-classical models outperformed classical methods in key areas, particularly in recursive architectures. The quantum models were also found to converge faster, meaning they required less time to train, which could lead to faster deployment in real-world applications.

The quantum-enhanced models showed competitive accuracy, especially as the number of qubits and the number of re-uploading blocks were increased. While purely classical models remain more computationally efficient, the quantum models demonstrated a marked improvement in predictive accuracy as they grew more complex.

These results suggest that, with further advancements in quantum computing, the hybrid models could eventually outperform classical techniques, particularly in environments where large, complex datasets are involved.

Implications for the Broader Public: A Smarter, More Efficient Future

The implications of this research extend beyond the realm of academic curiosity. If these quantum-enhanced traffic forecasting models can be deployed in real-world Intelligent Transportation Systems (ITS), they could transform urban traffic management.

Cities around the world struggle with congestion, pollution, and inefficient traffic flow—issues that often result in lost time, higher fuel consumption, and negative environmental impacts. By improving the accuracy of traffic forecasts, cities could implement more effective solutions, reducing congestion, improving emergency response times, and enhancing the overall quality of life for residents.

Moreover, this research demonstrates the growing potential of Quantum Machine Learning to tackle real-world challenges. While quantum computing is still in its early stages, its ability to process vast and complex datasets could be harnessed for other applications beyond traffic forecasting, such as climate modeling, healthcare diagnostics, and financial prediction.

A Step Towards the Future of Transportation

This study by Schetakis and colleagues presents an exciting leap forward in the application of Quantum Machine Learning to urban traffic forecasting.

While the results show that quantum models currently face challenges in terms of computational efficiency compared to classical models, the increased predictive accuracy and training efficiency of the quantum-enhanced models offer promising potential for future improvements.

As quantum computing technology continues to advance, the integration of quantum methods into traffic management systems could become a cornerstone of smart city initiatives worldwide.

The ongoing development of hybrid quantum-classical architectures and techniques like data re-uploading underscores the evolving landscape of machine learning and quantum computing.

For urban planners, policymakers, and technologists, this research offers a glimpse into how Quantum Neural Networks could revolutionize the way we approach traffic forecasting—and, ultimately, the way we manage and experience urban transportation.

For readers interested in staying up-to-date with the latest innovations in transportation, we invite you to explore more of our news articles. From breakthroughs in AI to cutting-edge developments in alternative energy, our collection of news stories offers a comprehensive look at how science and technology are reshaping the automotive world.

Reference:

Schetakis, N., Bonfini, P., Alisoltani, N. et al. Quantum neural networks with data re-uploading for urban traffic time series forecasting. Sci Rep 15, 19400 (2025).