TrafficKAN-GCN: Graph Convolutional-based Kolmogorov-Arnold Network for Traffic Flow Optimization

TL;DR

TrafficKAN-GCN tackles urban traffic optimization in dynamic, large-scale networks by blending Kolmogorov-Arnold Networks (KAN) with Graph Convolutional Networks (GCN) to learn expressive, nonlinear mappings on spatiotemporal graphs. The approach replaces fixed activations with adaptive univariate transformations and models the network as a weighted graph with edge weights incorporating length, speed, congestion, and travel time, using a normalized adjacency for stable learning. Empirical results on Baltimore-area data show that KAN-GCN offers robustness to noisy data and disruptions (e.g., the Key Bridge collapse) and can redistribute traffic to reduce peak-hour delays by about 22%, though it incurs higher training cost relative to MLP-GCN. The work highlights the potential of hybrid, stakeholder-aware graph learning for real-time urban mobility, while outlining future directions including lightweight architectures and Transformer-based temporal modeling for improved long-term forecasting and deployment in large-scale systems.

Abstract

Urban traffic optimization is critical for improving transportation efficiency and alleviating congestion, particularly in large-scale dynamic networks. Traditional methods, such as Dijkstra's and Floyd's algorithms, provide effective solutions in static settings, but they struggle with the spatial-temporal complexity of real-world traffic flows. In this work, we propose TrafficKAN-GCN, a hybrid deep learning framework combining Kolmogorov-Arnold Networks (KAN) with Graph Convolutional Networks (GCN), designed to enhance urban traffic flow optimization. By integrating KAN's adaptive nonlinear function approximation with GCN's spatial graph learning capabilities, TrafficKAN-GCN captures both complex traffic patterns and topological dependencies. We evaluate the proposed framework using real-world traffic data from the Baltimore Metropolitan area. Compared with baseline models such as MLP-GCN, standard GCN, and Transformer-based approaches, TrafficKAN-GCN achieves competitive prediction accuracy while demonstrating improved robustness in handling noisy and irregular traffic data. Our experiments further highlight the framework's ability to redistribute traffic flow, mitigate congestion, and adapt to disruptive events, such as the Francis Scott Key Bridge collapse. This study contributes to the growing body of work on hybrid graph learning for intelligent transportation systems, highlighting the potential of combining KAN and GCN for real-time traffic optimization. Future work will focus on reducing computational overhead and integrating Transformer-based temporal modeling for enhanced long-term traffic prediction. The proposed TrafficKAN-GCN framework offers a promising direction for data-driven urban mobility management, balancing predictive accuracy, robustness, and computational efficiency.

Figures (14)

• Figure 1: Baltimore's transportation network, showing major roads and highways (Left). The Francis Scott Key Bridge collapse on March 26, 2024, which disrupted traffic and economic activities (Right).
• Figure 2: Traffic Network Data with Road Characteristics and Weights
• Figure 3: Traffic Network Adjacency Matrix Heatmap: This heatmap visualizes the adjacency matrix, with darker colors indicating stronger connections.
• Figure 4: Roadmap of Tourists' Preferences
• Figure 5: Roadmap of Travelers' Preferences