Arxiv 2512.13745

A Spatio-Temporal Hybrid Quantum-Classical Graph Convolutional Neural Network Approach for Urban Taxi Destination Prediction

Xiuying Zhang , Qinsheng Zhu , and Xiaodong Xing

Abstract—We propose a Hybrid Spatio-Temporal Quantum Graph Convolutional Network (H-STQGCN) algorithm by combining the strengths of quantum computing and classical deep learning to predict the taxi destination within urban road networks. Our algorithm consists of two branches: spatial processing and time evolution. Regarding the spatial processing, the classical module encodes the local topological features of the road network based on the GCN method, and the quantum module is designed to map graph features onto parameterized quantum circuits through a differentiable pooling layer. The time evolution is solved by integrating multi-source contextual information and capturing dynamic trip dependencies on the classical TCN theory. Finally, our experimental results demonstrate that the proposed algorithm outperforms the current methods in terms of prediction accuracy and stability, validating the unique advantages of the quantum-enhanced mechanism in capturing high-dimensional spatial dependencies.

Index Terms—Intelligent transportation systems, vehicle destination prediction, quantum artificial intelligence, quantum graph convolutional network, parametric quantum circuit.

I. INTRODUCTION ith technological advancements and the increasing ownership of vehicles, the pressure on public transportation has intensified, particularly in megacities with high population densities. The challenge is obvious: rapid urbanization and growing traffic congestion exacerbate the problem. In this context, accurate taxi destination prediction is vital. It improves vehicle dispatch system efficiency, minimizes parking search time, and contributes to the optimization of urban planning and infrastructure [1], [2], [3].
Destination prediction research has developed from early statistical methods, Markov chains, and traditional machine

This research was funded by the Natural Science Foundation of Xinjiang Uygur Autonomous Region (No. 2024D01A17), and the Chengdu Key Research and Development Program (No. 2025-YF08-00109-GX). (Corresponding author: Qinsheng Zhu.) Xiuying Zhang is with the School of Physics, University of Electronic Science and Technology of China, Chengdu 610054, China (e-mail: zxy02402@163.com). Qinsheng Zhu is with the School of Physics, University of Electronic Science and Technology of China, Cheng Du, 610054, China and the Institute of Electronics and Information Industry Technology of Kash, Kash, 844000, China (e-mail: zhuqinsheng@uestc.edu.cn). Xiaodong Xing is with the School of Quantum Information Future Technology, Henan University, Zhengzhou 450046, China, the Henan Key Laboratory of Quantum Materials and Quantum Energy, Henan University, Zhengzhou 450046, China, and the Institute of Quantum Materials and Physics, Henan Academy of Sciences, Zhengzhou, 450046, China (e-mail: xiaodong.xing@henu.edu.cn). learning techniques that relied on shallow features such as speed and direction [4], [5], [6]. However, as trajectory data expands in scale and complexity, these shallow models fail to effectively learn the underlying complex patterns. This shift has led to the adoption of deep learning models, such as long short-term memory (LSTM) networks and temporal convolutional networks (TCN) [7], [8], [9]. These advancements have enhanced trajectory temporal modeling through stronger representation capabilities. Unfortunately, it struggles to effectively handle static road network topology and multi-level spatial dependencies, often compressing spatial features into a single vector [10], [11], [12]. To solve this problem, graph convolutional networks (GCN) and graph neural networks (GNN) have been widely adopted to aggregate neighborhood information via graph structures, thereby explicitly modeling spatial correlations in traffic prediction and path planning [8], [10], [13], [14]. Whereas, existing graph methods still encounter considerable limitations when dealing with high-dimensional, sparse, and dynamically evolving urban trajectory data, particularly regarding model depth, parameter efficiency, and dynamic spatiotemporal feature fusion [15], [16]. To overcome the limitations of classical computing in feature extraction, quantum computing offers a new approach in processing high-dimensional data. Based on the superposition and entanglement properties of qubits, quantum algorithms can map low-dimensional graph data into an exponentially dimensional Hilbert space, thereby capturing deep correlations and high-dimensional features that remain imperceptible to classical methods [4], [14], [19]. The proposal of hybrid quantum-classical neural network architectures [21] integrates parameterized quantum circuits (PQC) int

…(Full text truncated)…