Sanchita Basak, Abhishek Dubey, Bruno P. Leao
IEEE Big Data 2019
On-time performance is critical for public transit system reliability and user satisfaction, yet most transit agencies use manual approaches rather than systematic optimization. This work is innovative because it applies evolutionary and swarm-based optimization approaches to transit scheduling while accounting for seasonal and monthly variations in delay patterns. The clustering mechanism enables efficient optimization for multiple seasonal scenarios.
This paper presents data-driven optimization methods for public transit schedule optimization using genetic algorithms and particle swarm optimization. The approach addresses the challenge of optimizing bus timetables to maximize probability of on-time arrivals at timepoints within desired delay ranges. The work includes unsupervised clustering mechanisms for grouping months with similar seasonal delay patterns.
The genetic algorithm and particle swarm optimization approaches outperformed greedy baseline methods in improving on-time performance across Nashville bus routes. The particle swarm approach demonstrated faster convergence with lower variance compared to genetic algorithms. The system successfully optimized bus schedules for different seasonal patterns, improving overall system on-time performance.
@inproceedings{basak2019bigdata,
author = {Basak, Sanchita and Dubey, Abhishek and Leao, Bruno P.},
booktitle = {IEEE Big Data},
title = {Analyzing the Cascading Effect of Traffic Congestion Using LSTM Networks},
year = {2019},
address = {Los Angeles, Ca},
abstract = {This paper presents a data-driven approach for predicting the propagation of traffic congestion at road seg-ments as a function of the congestion in their neighboring segments. In the past, this problem has mostly been addressed by modelling the traffic congestion over some standard physical phenomenon through which it is difficult to capture all the modalities of such a dynamic and complex system. While other recent works have focused on applying a generalized data-driven technique on the whole network at once, they often ignore intersection characteristics. On the contrary, we propose a city-wide ensemble of intersection level connected LSTM models and propose mechanisms for identifying congestion events using the predictions from the networks. To reduce the search space of likely congestion sinks we use the likelihood of congestion propagation in neighboring road segments of a congestion source that we learn from the past historical data. We validated our congestion forecasting framework on the real world traffic data of Nashville, USA and identified the onset of congestion in each of the neighboring segments of any congestion source with an average precision of 0.9269 and an average recall of 0.9118 tested over ten congestion events.},
category = {selectiveconference},
contribution = {lead},
keywords = {transit scheduling, optimization, genetic algorithms, particle swarm optimization, schedule timetabling, seasonal variation},
tag = {ai4cps,incident,transit}
}