⚛️ quantum physics

QNN-VRCS: A Quantum Neural Network for Vehicle Road Cooperation Systems

This paper proposes QNN-VRCS, a quantum neural network integrating specific quantum algorithms and image encoding techniques to optimize Vehicle Road Cooperation Systems, achieving high classification accuracy and robustness in traffic data processing for intelligent 6G transportation solutions.

Original authors: Nouhaila Innan, Bikash K. Behera, Saif Al-Kuwari, Ahmed Farouk

Published 2026-10-26✓ Author reviewed ⓘ

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Nouhaila Innan, Bikash K. Behera, Saif Al-Kuwari, Ahmed Farouk

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a city where traffic lights do not just switch on a timer, but actually "see" the cars, understand the flow of the crowd, and make split-second decisions to keep everything moving smoothly. This vision of a smart, cooperative road system is the goal of modern transportation engineering, yet it faces a massive hurdle: the sheer volume of data. Every second, sensors and cameras generate a flood of information about vehicle positions, speeds, and light states. Traditional computers, which power our current traffic systems, often struggle to process this deluge quickly and accurately enough to prevent jams or accidents. They rely on established rules and patterns, but real-world traffic is messy, unpredictable, and constantly changing. To solve this, researchers are looking toward a new kind of computing power that operates on the strange and powerful laws of quantum mechanics, a field that promises to handle complex problems far beyond the reach of today's machines.

In a recent study, a team of scientists set out to test whether these quantum computers could improve how we manage traffic. They focused on a specific, critical task: teaching a computer to recognize traffic lights. The goal was to distinguish between red, yellow, and green signals with high accuracy, a fundamental requirement for any system that hopes to support future Vehicle Road Cooperation Systems. The researchers built a new type of artificial intelligence called a Quantum Neural Network. Unlike the neural networks found in standard smartphones or laptops, which process information in a linear, step-by-step fashion, this new model uses the unique properties of quantum bits, or qubits. These qubits can exist in multiple states at once, allowing the system to explore many possibilities simultaneously. The team combined this quantum power with advanced methods for turning images of traffic lights into quantum data, creating a system designed to learn and classify signals more reliably than current technology.

To see if their idea worked, the researchers put their Quantum Neural Network to the test against two different sets of traffic light images. The first set came from a computer simulation of a city, offering a clean and controlled environment. The second set was a collection of real-world photos taken from actual streets, which included the messy reality of changing light, shadows, and varying weather. They compared their quantum model against standard classical computer models and other experimental quantum methods. The results were striking. On the simulated city data, the quantum model correctly identified the traffic light color 97.42% of the time. On the more difficult, real-world photos, it still achieved an accuracy of 84.08%. In both cases, the quantum approach outperformed the traditional computer models, which struggled to reach the same level of precision, especially when the images were small or the conditions were complex.

The study did not just look at how well the model worked in a perfect world; it also tested how the system held up when things went wrong. Real quantum computers are currently very sensitive to their environment, often making mistakes due to tiny fluctuations in temperature or electromagnetic interference, known as noise. The researchers used simulated quantum noise models to investigate the system's robustness, testing six different types of these errors to see if their model would collapse. They found that while some types of noise, like those that cause energy to leak out of the system, did hurt performance, the model was surprisingly resilient to others. It maintained high accuracy even when the data was subjected to certain kinds of interference, suggesting that the system could potentially be robust under these simulated conditions.

This research offers a glimpse into a future where transportation systems are not just automated, but truly intelligent. By proving that a quantum neural network can classify traffic signals with superior accuracy relative to the specific classical and quantum models evaluated, the study suggests a path forward for 6G networks and autonomous vehicles that need to make instant, life-saving decisions. The work demonstrates that while the technology is still in its early stages and faces challenges like integrating with existing infrastructure, the potential for quantum computing to assist in how we move through our cities is becoming a tangible reality. The findings indicate that with further development, these quantum-enhanced systems could contribute to a new era in traffic management, turning the chaotic flow of urban life into a smooth, coordinated journey.

Technical Summary: QNN-VRCS: A Quantum Neural Network for Vehicle Road Cooperation Systems

Problem Overview
Urban transportation systems face escalating complexity due to traffic congestion, diverse modalities, and shifting commuter preferences. Traditional computational approaches struggle to process the voluminous datasets generated by real-time sensor networks, often lacking the precision required for accurate traffic prediction and system optimization. While Vehicle Road Cooperation Systems (VRCS) and autonomous vehicles rely on sophisticated visual processing for navigation and obstacle detection, current heuristic methods and classical machine learning models face limitations in handling unpredictable traffic dynamics and ensuring reliability across diverse scenarios. This research addresses the need for more sophisticated analytical frameworks by integrating quantum computing techniques to enhance VRCS, specifically focusing on the classification of traffic light states (red, yellow, green) to support future traffic management decisions.

Methodology
The study proposes a Quantum Neural Network (QNN) designed to classify traffic light images, comparing its performance against classical Convolutional Neural Networks (CNNs) and other quantum machine learning (QML) approaches.

Data Encoding: The research utilizes two primary quantum image representation methods:
- Flexible Representation of Quantum Images (FRQI): Encodes pixel values as rotation angles in a quantum state. For an image of size $2^n \times 2^n$ , this method requires $n+1$ qubits (e.g., 3 qubits for a $2 \times 2$ input).
- Novel Enhanced Quantum Representation (NEQR): Uses a series of qubits and CNOT gates to represent pixel intensity values directly, requiring more qubits (e.g., 10 qubits for a $2 \times 2$ input and 12 qubits for a $4 \times 4$ input) but preserving intensity information more explicitly.
- Angle Encoding (AE): Specifically used for the proposed QNN, where pixel values are converted to angles ( $\theta$ ) ranging from 0 to $\pi$ based on the formula $\theta = (\pi/255) \times \text{pixel value}$ .
Algorithms Evaluated:
1. $UU^\dagger$ Algorithm: A classification method where centroid data for each class (red, green, yellow) is encoded into a unitary operator $U$ , and test data is encoded into $U^\dagger$ . Classification is determined by measuring the inner product (probability of state $|00\rangle$ ) between the test state and the centroid states.
2. Variational $UU^\dagger$ Algorithm: Similar to the above but incorporates Hadamard gates before and after the measurement phase and utilizes repeating layers of $U$ and $U^\dagger$ gates to increase circuit depth.
3. Quantum Neural Network (QNN): A variational quantum circuit (VQC) approach. The model initializes qubits in $|0\rangle^{\otimes n}$ , encodes pixel values using rotational X operators, applies entangled layers using rotational Y and CNOT gates, and measures the output. The parameters are updated iteratively using a classical optimizer (Adam) to minimize the loss function (Mean Squared Error) between predicted and actual labels.
Experimental Setup:
- Datasets: Two datasets were used: the Carla dataset (1,909 images of traffic lights in a simulated environment) and the Cropped Lisa dataset (36,534 real-world images).
- Preprocessing: Images were resized to $2 \times 2$ and $4 \times 4$ pixels. A pre-trained TensorFlow model was used for object detection to isolate traffic lights.
- Noise Simulation: The QNN's robustness was evaluated against six simulated quantum noise models: bitflip, phaseflip, bitphaseflip, depolarizing, amplitude damping, and phase damping.

Key Contributions

Development of a QNN for VRCS: The paper presents the design and implementation of a QNN specifically tailored for vehicle road cooperation systems, focusing on traffic light classification.
Comparative Analysis: The study provides a comprehensive comparison between classical CNNs, $UU^\dagger$ , variational $UU^\dagger$ , and the proposed QNN across two distinct datasets and varying image resolutions.
Robustness Evaluation: The research assesses the resilience of the QNN model under six different simulated quantum noise conditions, offering insights into its stability in simulated noisy quantum environments.
Optimization of Quantum Encodings: The work explores the efficacy of different encoding strategies (FRQI, NEQR, and Angle Encoding) in conjunction with variational quantum algorithms for image classification tasks.

Results

Carla Dataset: The QNN with Angle Encoding and $2 \times 2$ pixel inputs achieved a high classification accuracy of 97.42%, outperforming the classical CNN (90% with $4 \times 4$ inputs) and other quantum methods (e.g., $UU^\dagger$ with FRQI achieved 66.25%) evaluated in this study. The QNN showed a clear trend of improving accuracy with increased layer depth, plateauing at 20 layers.
Cropped Lisa Dataset: The QNN with Angle Encoding and $2 \times 2$ inputs achieved a high accuracy of 84.08%, surpassing the classical CNN (77.09% with $2 \times 2$ inputs) and other quantum approaches. Variational $UU^\dagger$ models generally showed lower performance (around 43%) on this dataset.
Noise Resilience:
- The model demonstrated high stability against phase damping noise, maintaining high accuracy even at elevated noise intensities under simulated conditions.
- Amplitude damping noise caused a significant drop in performance, with accuracy falling to zero as the noise parameter exceeded 0.7.
- Depolarizing noise resulted in a steady decline in accuracy.
- The model showed higher overall accuracy on the Cropped Lisa dataset compared to Carla across all noise types, suggesting the real-world dataset may contain more distinguishable features.

Significance and Claims
The paper demonstrates that the proposed QNN model achieves improved traffic-light classification accuracy relative to the specific classical and quantum models evaluated. By achieving high classification accuracies (97.42% and 84.08%) with reduced image dimensions and minimal training data, the study highlights the potential of QNNs to serve as a component in future Intelligent Transportation Systems (ITS) and Vehicle Road Cooperation Systems. The research underscores the role of quantum technologies in advancing intelligent transportation solutions, particularly in the context of 6G-enabled systems.

The authors note that while the QNN demonstrates robustness under simulated noise conditions, challenges remain regarding the integration of quantum systems into existing classical infrastructure, latency issues for real-time decision-making, and data privacy. The study concludes that the QNN model provides a promising approach for traffic light classification tasks within VRCS, though future work must address the scalability of quantum circuits, the transition from simulated to physical quantum hardware, and hardware efficiency for practical deployment.

More like this