PAtt: A Pattern Attention Network for ETA Prediction Using Historical Speed Profiles

TL;DR

The paper tackles ETA prediction under dynamic traffic by introducing Pattern Attention (PAtt), a lightweight model that attends over historical pattern speed profiles to capture spatio-temporal causality along a route. It encodes three feature streams (background, per-link context, and pattern speeds), and refines temporal embeddings through k-cycle pattern attention to produce per-link duration estimates that aggregate into total travel time. Multi-cycle supervision combines per-link and route-level losses, leading to improved accuracy (MAPE of 8.78% on a large-scale real-world dataset) compared with rule-based baselines and neural baselines. The approach demonstrates robustness across route lengths and offers a scalable alternative to graph-based models for ETA prediction, with potential for adaptive pattern extraction and integration of real-time local events in future work.

Abstract

In this paper, we propose an ETA model (Estimated Time of Arrival) that leverages an attention mechanism over historical road speed patterns. As autonomous driving and intelligent transportation systems become increasingly prevalent, the need for accurate and reliable ETA estimation has grown, playing a vital role in navigation, mobility planning, and traffic management. However, predicting ETA remains a challenging task due to the dynamic and complex nature of traffic flow. Traditional methods often combine real-time and historical traffic data in simplistic ways, or rely on complex rule-based computations. While recent deep learning models have shown potential, they often require high computational costs and do not effectively capture the spatio-temporal patterns crucial for ETA prediction. ETA prediction inherently involves spatio-temporal causality, and our proposed model addresses this by leveraging attention mechanisms to extract and utilize temporal features accumulated at each spatio-temporal point along a route. This architecture enables efficient and accurate ETA estimation while keeping the model lightweight and scalable. We validate our approach using real-world driving datasets and demonstrate that our approach outperforms existing baselines by effectively integrating road characteristics, real-time traffic conditions, and historical speed patterns in a task-aware manner.

Figures (3)

• Figure 1: The overall architecture of Model: Contextual representation are extracted from background, link context via MLP layers, combined with temporal embedding, and processed through a cross attention with pattern speed profile and an estimation layer to produce temporal embedding. The procedure is repeated k times and the final temporal embedding is used to produce the final link-level ETA predictions.
• Figure 2: Performance comparison of loss functions with varying number of cycles
• Figure 3: Performance variation with respect to route length