On Robustness of the Generalized Proportional Controller for Traffic Signal Control

Gustav Nilsson, Giacomo Como

Research output: Chapter in Book/Report/Conference proceedingPaper in conference proceedingResearchpeer-review


We investigate robustness properties of the Generalized Proportional Allocation (GPA) policy that has been recently proposed for traffic signal control in urban networks. The GPA policy is fully decentralized, relies only on local information on the current congestion state, and requires no knowledge of the routing, the exogenous inflows, or the network structure. In previous work, we proved throughput optimality of the GPA policy, by showing that it is able to stabilize the traffic network dynamics whenever any controller is able to do so. In this paper, we first extend these results by showing that even when the measurements have offsets, the GPA policy maintains the same stability properties as with exact measurements. A comparison between the GPA and the MaxPressure traffic signal controllers with respect to robustness is also performed in a microscopic traffic simulator, where it is shown that while the GPA can handle offsets in the measurements, the MaxPressure controller may make vehicles wait forever.

Original languageEnglish
Title of host publication2020 American Control Conference, ACC 2020
PublisherInstitute of Electrical and Electronics Engineers Inc.
Number of pages6
ISBN (Electronic)9781538682661
Publication statusPublished - 2020 Jul
Externally publishedYes
Event2020 American Control Conference, ACC 2020 - Denver, United States
Duration: 2020 Jul 12020 Jul 3

Publication series

NameProceedings of the American Control Conference
ISSN (Print)0743-1619


Conference2020 American Control Conference, ACC 2020
Country/TerritoryUnited States

Subject classification (UKÄ)

  • Control Engineering


  • Decentralized Control
  • Traffic Signal Control
  • Transportation Networks


Dive into the research topics of 'On Robustness of the Generalized Proportional Controller for Traffic Signal Control'. Together they form a unique fingerprint.

Cite this