Online Equitable Ride-sharing Car Distribution in a Congested Traffic

TL;DR

This paper proposes a novel traffic dynamics model to promote equitable distribution of ride-sharing cars in congested urban networks, aiming to reduce congestion and improve service fairness.

Contribution

It introduces a dissensus-based traffic evolution model using non-stationary Markov processes with decision variables for equitable ride-sharing car distribution.

Findings

Model effectively captures ride-sharing car dynamics in congested networks.

Decision variables can be optimized for equitable distribution.

Potential to reduce congestion and improve service fairness.

Abstract

The usability of ride-sharing services like Uber and Lyft has been considerably improved by advancements in cellular communications. Such a tech-driven transportation system can reduce the number of private cars, in roads with limited physical capacity, effectively match drivers with passengers requesting service, and save drivers and passengers time by optimal scheduling and estimating of the trips. However, the existing services offered by ride-sharing companies may not necessarily reduce congestion, if they are not equitably accessed in urban areas. This paper addresses this important issue by classifying cars as ride-sharing and non-ride-sharing cars and developing a novel dissensus-based traffic evolution dynamics to effectively model their distribution in a shared network of interconnected roads (NOIR). This new dynamics models the ride-sharing car evolution as a non-stationary Markov process with uncertainty matrices decomposed into outflow and tendency probability matrices, where outflow probabilities are constrained by the non-ride-sharing cars behavior, but the tendency probabilities are considered as decision variables (or distributed controls) for achieving an equitable distribution of ride-sharing cars in a congested traffic.