Penalization and augmented Lagrangian for O-D demand matrix estimation from transit segment counts
place - north america, planning - methods, ridership - demand
O-D matrix, transit assignment, penalization, augmented Lagrangian, conjugate gradient method
In this paper we consider some penalized quadratic models to update Origin-Destination (O-D) matrices in transit networks from observed flows. These models look for the closest O-D matrix to an outdated one, which reproduces some observed segment flows. We demonstrate that the solution of these penalized models converges to the solution of the Spiess model when the penalty parameter increases to infinity. Another contribution is the introduction of an augmented Lagrangian model and its iterative solution by a dual ascent technique and the method of multipliers. This approach yields high-quality solutions with low CPU time and it is tested with two networks: the Winnipeg transit network, which has 23716 O-D pairs; and the transit network of the metropolitan area of the Valley of Mexico with more than 2 million of O-D pairs. For some instances, extracting the null coefficients from the old O-D matrix reduces the computational cost even further.
Permission to publish the abstract has been given by Taylor&Francis, copyright remains with them.
Chávez Hernández, M.V., Juárez Valencia, L.H., & Ríos Solís, Y.A. (2019). Penalization and augmented Lagrangian for O-D demand matrix estimation from transit segment counts. Transportmetrica A: Transport Science, Vol. 15, pp. 915-943.