Traffic dynamics and optimal control in a city served by ride-sourcing vehicles

TL;DR

This paper introduces an interactive bathtub model for ride-sourcing traffic dynamics, enabling effective traffic control and ride-pooling management in a city with ride-sourcing vehicles, using minimal input data.

Contribution

It develops a parsimonious, dynamic model capturing traffic and passenger states, and proposes robust control strategies for traffic management and ride-pooling optimization.

Findings

Model accurately captures traffic and passenger dynamics.

Control strategies effectively prevent gridlock.

Ride-pooling sizes can be optimized to reduce unmatched requests.

Abstract

This paper presents an interactive bathtub model for describing the traffic dynamics of ride-sourcing vehicles including non-shared taxis and ride-pooling cars. A city with a network of undifferentiated streets and solely served by ride-sourcing services is assumed to facilitate the modeling, isolate the congestion contribution, and accordingly develop control strategies. The proposed model is parsimonious with only input information of the total lane length of the network, the in-flux of demand, and the travel distance distributions. The output of the model, however, captures not only the traffic dynamics of vehicles but also the dynamic states of passengers in ride-pooling services in terms of the total number, the remaining travel distances, and the queue of unmatched requests at any system time. Useful system metrics can be exploited for use of the authorities to monitor, predict, and control the traffic, as well as for the TNCs to determine the fleet sizes, dispatch vehicles, and measure the service productivity. For illustration, we propose a robust control rule to manage the traffic efficiently and avoid gridlock, and also present time-varying ride-pooling sizes to eliminate the queue of unmatched requests. Numerical examples demonstrate the effectiveness of the proposed model and the control and operation strategies.