Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Public Transit Arrival Prediction: a Seq2Seq RNN Approach

Nancy Bhutani, Soumen Pachal, Avinash Achar

TL;DR

The paper tackles bus arrival time prediction under challenging mixed-traffic conditions typical of developing regions, proposing a novel Encoder-Decoder (Seq2Seq) RNN built on GRU units. It introduces a bidirectional decoder and a space-time-synced input scheme that leverages current bus travel times, the closest previous bus, and a previous-week seasonal trip to model spatio-temporal and seasonal correlations without time discretization. Empirical results on a Delhi route show that the bidirectional ED variant (EDB) consistently outperforms multiple strong baselines (LNKF, SVKF, DpAR, CLSTM, and EDU) across short- and long-horizon predictions, with statistically significant improvements. The approach offers a flexible, real-time forecasting framework that can be extended to other domains and future work includes transformer-based variants and additional exogenous inputs.

Abstract

Arrival/Travel times for public transit exhibit variability on account of factors like seasonality, dwell times at bus stops, traffic signals, travel demand fluctuation etc. The developing world in particular is plagued by additional factors like lack of lane discipline, excess vehicles, diverse modes of transport and so on. This renders the bus arrival time prediction (BATP) to be a challenging problem especially in the developing world. A novel data-driven model based on recurrent neural networks (RNNs) is proposed for BATP (in real-time) in the current work. The model intelligently incorporates both spatial and temporal correlations in a unique (non-linear) fashion distinct from existing approaches. In particular, we propose a Gated Recurrent Unit (GRU) based Encoder-Decoder(ED) OR Seq2Seq RNN model (originally introduced for language translation) for BATP. The geometry of the dynamic real time BATP problem enables a nice fit with the Encoder-Decoder based RNN structure. We feed relevant additional synchronized inputs (from previous trips) at each step of the decoder (a feature classically unexplored in machine translation applications). Further motivated from accurately modelling congestion influences on travel time prediction, we additionally propose to use a bidirectional layer at the decoder (something unexplored in other time-series based ED application contexts). The effectiveness of the proposed algorithms is demonstrated on real field data collected from challenging traffic conditions. Our experiments indicate that the proposed method outperforms diverse existing state-of-art data-driven approaches proposed for the same problem.

Public Transit Arrival Prediction: a Seq2Seq RNN Approach

TL;DR

The paper tackles bus arrival time prediction under challenging mixed-traffic conditions typical of developing regions, proposing a novel Encoder-Decoder (Seq2Seq) RNN built on GRU units. It introduces a bidirectional decoder and a space-time-synced input scheme that leverages current bus travel times, the closest previous bus, and a previous-week seasonal trip to model spatio-temporal and seasonal correlations without time discretization. Empirical results on a Delhi route show that the bidirectional ED variant (EDB) consistently outperforms multiple strong baselines (LNKF, SVKF, DpAR, CLSTM, and EDU) across short- and long-horizon predictions, with statistically significant improvements. The approach offers a flexible, real-time forecasting framework that can be extended to other domains and future work includes transformer-based variants and additional exogenous inputs.

Abstract

Arrival/Travel times for public transit exhibit variability on account of factors like seasonality, dwell times at bus stops, traffic signals, travel demand fluctuation etc. The developing world in particular is plagued by additional factors like lack of lane discipline, excess vehicles, diverse modes of transport and so on. This renders the bus arrival time prediction (BATP) to be a challenging problem especially in the developing world. A novel data-driven model based on recurrent neural networks (RNNs) is proposed for BATP (in real-time) in the current work. The model intelligently incorporates both spatial and temporal correlations in a unique (non-linear) fashion distinct from existing approaches. In particular, we propose a Gated Recurrent Unit (GRU) based Encoder-Decoder(ED) OR Seq2Seq RNN model (originally introduced for language translation) for BATP. The geometry of the dynamic real time BATP problem enables a nice fit with the Encoder-Decoder based RNN structure. We feed relevant additional synchronized inputs (from previous trips) at each step of the decoder (a feature classically unexplored in machine translation applications). Further motivated from accurately modelling congestion influences on travel time prediction, we additionally propose to use a bidirectional layer at the decoder (something unexplored in other time-series based ED application contexts). The effectiveness of the proposed algorithms is demonstrated on real field data collected from challenging traffic conditions. Our experiments indicate that the proposed method outperforms diverse existing state-of-art data-driven approaches proposed for the same problem.
Paper Structure (20 sections, 7 equations, 7 figures, 1 table)

This paper contains 20 sections, 7 equations, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. Literature Review and Related Work
  3. Data-driven methods for BATP
  4. Related ED based RNN approaches
  5. Proposed approach in perspective of related approaches
  6. Methodology
  7. Data Preliminaries
  8. Capturing Spatio-Temporal and Seasonal Correlations
  9. Dynamic Real-time Prediction Problem
  10. Motivating the proposed ED Variant Architecture
  11. ED Variant with Bidirectional Decoder
  12. Gating Unit in the architecture
  13. Bidirectional Layer at the Decoder
  14. Training Data Preparation
  15. Results
  16. ...and 5 more sections

Figures (7)

  • Figure 1: Spatio-temporal and Seasonal Correlations pictorially. Current Bus Inputs (Spatial correlations), Previous Bus Inputs (Temporal Correlations), Previous Week Bus Inputs (Weekly Seasonal Correlations).
  • Figure 2: Proposed ED Architecture with a Unidirectional Decoder. $K=(N_s - m)$
  • Figure 3: Bidirectional Decoder. Please note $K=(N_s - m)$. $h_{K+1}^\leftarrow$ and $h_0^\rightarrow$ are set to $E_a$, output of the append block in Fig. \ref{['fig:EDArchitecture']}.
  • Figure 4: Two-Step Ahead MAPE and MAE Comparison at a Day level
  • Figure 5: Multi-Step Ahead MAPE at various start/current bus positions (again chosen in steps of 5) along the route.
  • ...and 2 more figures

Theorems & Definitions (3)

  • Remark 1
  • Remark 2
  • Remark 3