Compressive origin-destination estimation

Abstract

The paper presents an approach to estimate Origin-Destination (OD) flows and their path splits, based on traffic counts on links in the network. The approach called Compressive Origin-Destination Estimation (CODE) is inspired by Compressive Sensing (CS) techniques. Even though the estimation problem is underdetermined, CODE recovers the unknown variables exactly when the number of alternative paths for each OD pair is small. Noiseless, noisy, and weighted versions of CODE are illustrated for synthetic networks, and with real data for a small region in East Providence. CODE's versatility is suggested by its use to estimate the number of vehicles and the Vehicle-Miles Traveled (VMT) using link counts.

Keywords:

• Origin-destination matrix estimation
• VMT estimation
• Compressive sensing
• ℓ₁ Minimization
• Link traffic counts