ST-Mamba: Spatial-Temporal Mamba for Traffic Flow Estimation Recovery using Limited Data

TL;DR

Real-time traffic flow estimation often suffers from incomplete data when relying on vehicular networks. ST-Mamba merges CNN-based spatial feature learning with a Mamba-based temporal model in an encoder–Mamba–decoder framework on a grid representation, learning from history of limited data to approximate $Z_t$ from $[\boldsymbol{X}_{t-L+1}, \dots, \boldsymbol{X}_t]$ via $\mathcal{F}$. The approach introduces a spatial-temporal fusion that leverages per-grid CNN features and a continuous-time state-space–inspired Mamba with an input-dependent selection mechanism, achieving accurate and stable city-wide TFE under data scarcity. Validated on Beijing data across $10\%$ to $50\%$ data availability, ST-Mamba outperforms baselines, offering a cost-effective, scalable solution for urban traffic management with reduced data collection and communication overhead.

Abstract

Traffic flow estimation (TFE) is crucial for urban intelligent traffic systems. While traditional on-road detectors are hindered by limited coverage and high costs, cloud computing and data mining of vehicular network data, such as driving speeds and GPS coordinates, present a promising and cost-effective alternative. Furthermore, minimizing data collection can significantly reduce overhead. However, limited data can lead to inaccuracies and instability in TFE. To address this, we introduce the spatial-temporal Mamba (ST-Mamba), a deep learning model combining a convolutional neural network (CNN) with a Mamba framework. ST-Mamba is designed to enhance TFE accuracy and stability by effectively capturing the spatial-temporal patterns within traffic flow. Our model aims to achieve results comparable to those from extensive data sets while only utilizing minimal data. Simulations using real-world datasets have validated our model's ability to deliver precise and stable TFE across an urban landscape based on limited data, establishing a cost-efficient solution for TFE.

Figures (6)

• Figure 1: An illustration of TFE with limited data.
• Figure 2: ST-Mamba framework.
• Figure 3: The ideal TFE on Friday 5:00 PM.
• Figure 4: The original TFE on Friday 5:00 PM at 10% limitation.
• Figure 5: The recovered TFE on Friday 5:00 PM at 10% limitation.