Data preconditioning for data-driven predictive control in urban traffic control


Loading...

Author / Producer

Date

2024-10-07

Publication Type

Master Thesis

ETH Bibliography

yes

Citations

Altmetric

Data

Abstract

Traffic congestion in urban areas presents a significant challenge, often without the possibility of expanding road infrastructure. To address this, we employ Data-EnablEd Predictive Control (DeePC) to dynamically adjust the green-red cycles of traffic lights based on current traffic conditions. Given the nonlinear nature of urban traffic networks, we propose modeling them as Linear Parameter-Varying (LPV) systems. Our research demonstrates how to adapt the data used by DeePC for decision-making, contingent on known switching times of the LPV system. We establish the equivalence of closed-loop behavior between Model-Predictive Control (MPC) and DeePC during specific intervals of the LPV system. Offline data clustering is utilized to group traffic data from various days into behavior groups, which represent the subsystems of our urban traffic network modeled as an LPV system. The most appropriate data cluster is selected online to match current traffic conditions, and the controller’s data is updated online as the urban traffic network transitions between different behaviors when we control the urban traffic network with our proposed data-switching DeePC algorithm. Numerical simulations illustrate that the proposed algorithm enhances the robustness of traffic congestion prevention and improves computational efficiency compared to a DeePC controller with fixed data.

Publication status

published

External links

Editor

Contributors

Examiner : Rimoldi, Alessio
Examiner : Cenedese, Carlo
Examiner: Padoan, Alberto
Examiner: Lygeros, John

Book title

Journal / series

Volume

Pages / Article No.

Publisher

ETH Zurich

Event

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

smart mobility; DeePC; Urban traffic control; LPV systems; Data clustering

Organisational unit

03751 - Lygeros, John / Lygeros, John check_circle

Notes

Funding

Related publications and datasets