Real Time Headway Predictions in Urban Rail Systems and Implications for Service Control: A Deep Learning Approach

TL;DR

Real-time dispatching in urban metro systems is essential for reliability and passenger satisfaction due to cascading headway effects along the line. The paper proposes a ConvLSTM-based spatiotemporal framework that jointly models headway propagation and ingests planned terminal headways to forecast downstream headways, accompanied by a data preprocessing pipeline that converts AVL trajectories into a space-time grid and a simulator of dispatcher strategies. Key contributions include direct terminal-headway inputs, scalable headway predictions across the full line, and demonstration on a large-scale simulated CTA Blue Line dataset with 15–60 minute horizons, showing strong short-term accuracy and useful longer-horizon performance for proactive control. The work provides rail operators with a computationally efficient tool to test and optimize terminal-control strategies, potentially improving service reliability and passenger experience.

Abstract

Efficient real-time dispatching in urban metro systems is essential for ensuring service reliability, maximizing resource utilization, and improving passenger satisfaction. This study presents a novel deep learning framework centered on a Convolutional Long Short-Term Memory (ConvLSTM) model designed to predict the complex spatiotemporal propagation of train headways across an entire metro line. By directly incorporating planned terminal headways as a critical input alongside historical headway data, the proposed model accurately forecasts future headway dynamics, effectively capturing both their temporal evolution and spatial dependencies across all stations. This capability empowers dispatchers to evaluate the impact of various terminal headway control decisions without resorting to computationally intensive simulations. We introduce a flexible methodology to simulate diverse dispatcher strategies, ranging from maintaining even headways to implementing custom patterns derived from observed terminal departures. In contrast to existing research primarily focused on passenger load predictioning or atypical disruption scenarios, our approach emphasizes proactive operational control. Evaluated on a large-scale dataset from an urban metro line, the proposed ConvLSTM model demonstrates promising headway predictions, offering actionable insights for real-time decision-making. This framework provides rail operators with a powerful, computationally efficient tool to optimize dispatching strategies, thereby significantly improving service consistency and passenger satisfaction.

Figures (9)

• Figure 1: Train trajectories are plotted with time ($t$) on the x-axis and distance ($d$) on the y-axis, illustrating the spatio-temporal locations of signal block activations with headways, discretized into grid cells of size $\Delta T \times \Delta d$.
• Figure 2: ConvLSTM-based model architecture for spatiotemporal headway prediction, integrating historical and planned terminal headway inputs.
• Figure 3: Map of CTA Blue Line with terminals at Forest Park and O'Hare and UIC-Halsted serve as short-turning station
• Figure 4: Headway heatmaps for the Blue Line during the afternoon peak period (15:30--18:00). The plots illustrate headway values (in seconds) across 64 distance bins and 1-minute time bins. The headway changes at distance bin 21 indicate the short-turning station UIC-Halsted, where southbound trains switch to northbound, leading to a complex operational situation.
• Figure 5: Training and Validation Loss over Epochs