Adaptive Traffic Signal Control: Game-Theoretic Decentralized vs. Centralized Perimeter Control

This paper compares the operation of a decentralized Nash bargaining traffic signal controller (DNB) to the operation of state-of-the-art adaptive and gating traffic signal control. Perimeter control (gating), based on the network fundamental diagram (NFD), was applied on the borders of a protected urban network (PN) to prevent and/or disperse traffic congestion. The operation of gating control and local adaptive controllers was compared to the operation of the developed DNB traffic signal controller. The controllers were implemented and their performance assessed on a grid network in the INTEGRATION microscopic simulation software. The results show that the DNB controller, although not designed to solve perimeter control problems, successfully prevents congestion from building inside the PN and improves the performance of the entire network. Specifically, the DNB controller outperforms both gating and non-gating controllers, with reductions in the average travel time ranging between $21\%$ and $41\%$, total delay ranging between $40\%$ and $55\%$, and emission levels/fuel consumption ranging between $12\%$ and $20\%$. The results demonstrate statistically significant benefits of using the developed DNB controller over other state-of-the-art centralized and decentralized gating/adaptive traffic signal controllers.