A new model for multi-commodity macroscopic modeling of complex traffic networks

TL;DR

This paper introduces the Link-Node Cell Transmission Model (LNCTM), a novel macroscopic framework for multi-commodity traffic networks that captures capacity drops, hysteresis, and dynamic routing at nodes.

Contribution

The paper develops the LNCTM, an advanced multi-commodity traffic model with a new multi-input-multi-output node model and a local traffic assignment algorithm, improving upon previous models.

Findings

The LNCTM effectively models capacity drop and hysteresis behaviors.

The multi-commodity node model outperforms previous models in accuracy.

The local traffic assignment algorithm dynamically distributes traffic based on current network states.

Abstract

We propose a macroscopic modeling framework for a network of roads and multi-commodity traffic. The proposed framework is based on the Lighthill-Whitham-Richards kinematic wave theory; more precisely, on its discretization, the Cell Transmission Model (CTM), adapted for networks and multi-commodity traffic. The resulting model is called the Link-Node CTM (LNCTM). In the LNCTM, we use the fundamental diagram of an "inverse lambda" shape that allows modeling of the capacity drop and the hysteresis behavior of the traffic state in a link that goes from free flow to congestion and back. A model of the node with multiple input and multiple output links accepting multi-commodity traffic is a cornerstone of the LNCTM. We present the multi-input-multi-output (MIMO) node model for multi-commodity traffic that supersedes previously developed node models. The analysis and comparison with previous node models are provided. Sometimes, certain traffic commodities may choose between multiple output links in a node based on the current traffic state of the node's input and output links. For such situations, we propose a local traffic assignment algorithm that computes how incoming traffic of a certain commodity should be distributed between output links, if this information is not known a priori.