Modeling vehicular mobility patterns using recurrent neural networks

TL;DR

This paper demonstrates that recurrent neural networks can effectively model and reproduce complex vehicular mobility patterns, including spatial distributions, outperforming other models in accuracy.

Contribution

The study introduces the use of recurrent neural networks for generating realistic vehicular mobility patterns, a task previously limited by scarce accurate generative models.

Findings

RNNs accurately reproduce spatial distributions of street segment usage

Models trained on NYC, Shanghai, and Michigan data show broad applicability

RNNs outperform existing models in capturing mobility patterns

Abstract

Data on vehicular mobility patterns have proved useful in many contexts. Yet generative models which accurately reproduce these mobility patterns are scarce. Here, we explore if recurrent neural networks can cure this scarcity. By training networks on taxi from NYC and Shanghai, and personal cars from Michigan, we show most aspects of the mobility patterns can be reproduced. In particular, the spatial distributions of the street segments usage is well captured by the recurrent neural networks, which other models struggle to do.