Data-Driven Model for Traffic Signal Control
Chen Zhang, Yugeng Xi, Dewei Li, Yunwen Xu
Abstract
Chen Zhang, Yugeng Xi, Dewei Li, Yunwen Xu
Abstract
With increasing traffic demand and limited transportation structure, traffic congestion is a global and severe problem. This paper proposes a novel data-driven model based control strategy. Traditionally the urban traffic control needs some traffic data such as traffic density and saturation, which is hard to collect in practice. This paper uses data of traffic volume to build a dynamic state transition model between real time traffic volume and density. This model based on Hidden Markov Model is used to predict traffic density, which reflects the level of traffic congestion directly. We also design a signal control framework of each intersection. The input of control system is historical traffic data and predictive density value provided by the traffic data model, and the output is optimal traffic timings of traffic phases allocation. The numerical experiments of a subnetwork with 19 intersections show that this data driven model based control strategy decreases the congestion effectively under medium and high traffic demand.
OpenAlex reports 5 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
With increasing traffic demand and limited transportation structure, traffic congestion is a global and severe problem. This paper proposes a novel data-driven model based control strategy. Traditionally the urban traffic control needs some traffic data such as traffic density and saturation, which is hard to collect in practice. This paper uses data of traffic volume to build a dynamic state transition model between real time traffic volume and density. This model based on Hidden Markov Model is used to predict traffic density, which reflects the level of traffic congestion directly. We also design a signal control framework of each intersection. The input of control system is historical traffic data and predictive density value provided by the traffic data model, and the output is optimal traffic timings of traffic phases allocation. The numerical experiments of a subnetwork with 19 intersections show that this data driven model based control strategy decreases the congestion effectively under medium and high traffic demand.
Key concepts: Traffic congestion reconstruction with Kerner's three-phase theory, Traffic generation model, Computer science, Floating car data, Intersection (aeronautics), Traffic congestion, Network traffic control, Data modeling