Ramp metering control under stochastic capacity in a connected environment: A dynamic bargaining game theory approach

• Published in: Transportation research. Part C, Emerging technologies Emerging technologies, 2021-09, Vol.130, p.103282, Article 103282
• Main Authors: Heshami, Seiran, Kattan, Lina
• Format: Article
• Language: English
• Subjects: Bargaining game; Connected vehicles; Ramp metering; Stochastic capacity
• ISSN: 0968-090X; 1879-2359
• DOI: 10.1016/j.trc.2021.103282

•Probabilistic breakdown in individual lanes, is embodied in the ramp metering method.•Stochasticity in demand and capacity is incorporated in the prediction and control.•The proposed bargaining game achieves a more equitable distribution of breakdown events, while seeking system-wide efficiency.•The performance of the scenarios are evaluated based on efficiency, effectiveness, and equity.•Effectiveness, as a novel performance measure, is evaluated based on the number of failed control actions.•A novel equity measure based on a more balanced distribution of the control failures is introduced.•A balanced trade off between equity and efficiency is achieved via proposed bargaining game. This paper presents a dynamic predictive and cooperative ramp metering approach that considers stochastic breakdowns at merging bottlenecks. A stochastic microscopic model is used to estimate traffic state parameters based on speed, location, and travel time information from connected vehicles. Traffic state predictions are obtained on a lane by lane basis using an adaptive Kalman filter (AKF) that fuses fixed detector measurements with the model; the AKF then produces multiple step ahead predictions. The ramp metering problem in this paper is modeled as a stochastic distributed model predictive control (SDMPC) approach. The SDMPC problem is solved based on a bargaining game approach where each controller, a player in the game, receives traffic state and control decision information from other controllers to solve the local optimization problem based on expected local costs and constraints. The performance of the proposed model is evaluated for three aspects of efficiency: short-term and long-term equity and effectiveness compared to multiple control scenarios. The outcomes indicate that the proposed cooperative model with stochastic capacity considerations outperforms the deterministic capacity-based models in regard to effectiveness and equity properties. However, the centralized approach performs slightly better in respect to system-wide efficiency.