Interpretable Cascading Mixture-of-Experts for Urban Traffic Congestion Prediction

TL;DR

CP-MoE introduces a scalable, interpretable congestion-prediction framework by cascading a sparsely gated mixture of adaptive graph learners (MAGLs) with trend and periodic experts (CITPE) and calibrating expert collaboration through ordinal regression (ECB). The model addresses heterogeneity, data noise, and missingness in urban traffic by using specialized graph experts (upstream/downstream/global) and a robust gating strategy that leverages rich gate inputs and confidence-based fusion. Empirical results on Beijing and Shanghai datasets show CP-MoE outperforms state-of-the-art spatio-temporal models, with notable robustness to corruption and clear interpretability via expert weights; deployment within DiDi improves real-world travel time estimation. The optimization combines an ordinal-regression loss with expert balancing terms, formulated as $\mathcal{L}=\mathcal{L}_{ord}+\lambda_1\sum_{l=1}^L\mathcal{L}_{imp}^{(l)}+\lambda_2\sum_{l=1}^L\mathcal{L}_{load}^{(l)}$, ensuring balanced, diverse expert collaboration across heterogeneous traffic scenarios.

Abstract

Rapid urbanization has significantly escalated traffic congestion, underscoring the need for advanced congestion prediction services to bolster intelligent transportation systems. As one of the world's largest ride-hailing platforms, DiDi places great emphasis on the accuracy of congestion prediction to enhance the effectiveness and reliability of their real-time services, such as travel time estimation and route planning. Despite numerous efforts have been made on congestion prediction, most of them fall short in handling heterogeneous and dynamic spatio-temporal dependencies (e.g., periodic and non-periodic congestions), particularly in the presence of noisy and incomplete traffic data. In this paper, we introduce a Congestion Prediction Mixture-of-Experts, CP-MoE, to address the above challenges. We first propose a sparsely-gated Mixture of Adaptive Graph Learners (MAGLs) with congestion-aware inductive biases to improve the model capacity for efficiently capturing complex spatio-temporal dependencies in varying traffic scenarios. Then, we devise two specialized experts to help identify stable trends and periodic patterns within the traffic data, respectively. By cascading these experts with MAGLs, CP-MoE delivers congestion predictions in a more robust and interpretable manner. Furthermore, an ordinal regression strategy is adopted to facilitate effective collaboration among diverse experts. Extensive experiments on real-world datasets demonstrate the superiority of our proposed method compared with state-of-the-art spatio-temporal prediction models. More importantly, CP-MoE has been deployed in DiDi to improve the accuracy and reliability of the travel time estimation system.

Figures (7)

• Figure 1: Primary data analysis on Beijing dataset. In (a) and (b), deep color indicates higher dependency. In (c), deeper color implies higher instability of traffic condition. In (d), deeper color implies a higher likelihood of periodic congestion.
• Figure 2: Overall framework of CP-MoE.
• Figure 3: Robustness check on the Beijing dataset.
• Figure 4: Ablation study on Beijing and Shanghai datasets.
• Figure 5: Expert weight distribution on Shanghai Dataset.

Theorems & Definitions (2)

• Definition 1: Traffic Network
• Definition 2: Congestion Level