Empirical validation of network learning with taxi GPS data from Wuhan, China

TL;DR

This paper empirically validates a cost-effective multi-agent inverse optimization method for monitoring transportation networks using taxi GPS data from Wuhan, demonstrating its ability to accurately estimate travel times with minimal data.

Contribution

The study provides real-world validation of a previously proposed low-cost network monitoring method using taxi GPS data from Wuhan, China.

Findings

Using data from one OD pair, the method achieves a 0.23 correlation with observed travel times.

Monitoring two OD pairs increases correlation to 0.56.

The approach effectively estimates network parameters with limited data.

Abstract

In prior research, a statistically cheap method was developed to monitor transportation network performance by using only a few groups of agents without having to forecast the population flows. The current study validates this "multi-agent inverse optimization" method using taxi GPS probe data from the city of Wuhan, China. Using a controlled 2062-link network environment and different GPS data processing algorithms, an online monitoring environment is simulated using the real data over a 4-hour period. Results show that using only samples from one OD pair, the multi-agent inverse optimization method can learn network parameters such that forecasted travel times have a 0.23 correlation with the observed travel times. By increasing to monitoring from just two OD pairs, the correlation improves further to 0.56.