Integrated Traffic Control for Freeway Recurrent Bottleneck Based on Deep Reinforcement Learning

Wang, Chong; Xu, Yang; Zhang, Jian

doi:10.1109/tits.2022.3141730

Cited by 21 publications

(11 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To quantitatively analyze the simulation results, this paper chooses Total Travel Time (TTT) [65,66] and Average Travel Speeds (ATS) [38,67] as the performance measures. Tables 3-5 present the TTT and ATS for various simulation scenarios under 20% PR, where each table corresponds to one demand split.…”

Section: Simulation Resultsmentioning

confidence: 99%

“…The frequently utilized macroscopic traffic flow control methods at merging areas are Variable Speed Limit (VSL) control [28–31], ramp metering [32–36], and integrated control [37, 38]. For instance, Ke et al.…”

Section: Related Workmentioning

confidence: 99%

“…Wang et al. [38] developed a centralized traffic control method that utilizes the deep reinforcement learning algorithm to coordinate multiple ramp metering and VSL traffic controllers on freeways. The most advanced aspect of this approach is that it is not affected by the curse of dimensionality problem in control systems.…”

Section: Related Workmentioning

confidence: 99%

“…The frequently utilized macroscopic traffic flow control methods at merging areas are Variable Speed Limit (VSL) control [28][29][30][31], ramp metering [32][33][34][35][36], and integrated control [37,38]. For instance, Ke et al [31] developed a VSL strategy based on the deep reinforcement learning algorithm and used transfer learning to improve the transferability of the control strategy.…”

Section: Macroscopic Traffic Flow Control At Freeway Merging Areasmentioning

confidence: 99%

“…Their proposed method can automatically adjust ramp metering and VSL based on real-time traffic conditions and bottleneck locations, thereby maximizing road capacity and traffic efficiency. Wang et al [38] developed a centralized traffic control method that utilizes the deep reinforcement learning algorithm to coordinate multiple ramp metering and VSL traffic controllers on freeways. The most advanced aspect of this approach is that it is not affected by the curse of dimensionality problem in control systems.…”

Section: Macroscopic Traffic Flow Control At Freeway Merging Areasmentioning

confidence: 99%

See 4 more Smart Citations

Multilane freeway merging control via trajectory optimization in a mixed traffic environment

Han

Zhang

Guo

2023

IET Intelligent Trans Sys

View full text Add to dashboard Cite

Merging activities at freeway merging areas can cause significant recurrent and non‐recurrent bottleneck congestion due to vehicles’ mandatory lateral conflicts. The Connected and Automated Vehicles (CAVs), with their capabilities of real‐time communication and precise trajectory control, hold great potential to prevent or mitigate such critical conflicts at merging areas. However, the performance of CAVs may be impaired by the imbalance of lane flow distribution, and non‐cooperative movements of Human‐driven Vehicles (HDVs) in the mixed traffic environment (i.e. traffic mixed with CAVs and HDVs). In this paper, a novel two‐level hierarchical traffic control framework for multilane merging areas under the mixed traffic environment is developed. Note that this paper assumes that the merging sequence is determined by the high control level and focuses on the low level of the control framework. The low control level not only establishes a trajectory optimization strategy for CAVs with lane‐changing optimization and cooperative merging control, but includes a human‐like merging strategy for HDVs. First, to balance downstream lane flow distribution and provide sufficient merging space for on‐ramp vehicles, a lane‐changing optimization method is proposed to choose a certain number of designated mainline CAVs to perform early lane changes at the upstream mainline. Second, a cooperative merging control method is presented to optimize the longitudinal trajectories of both mainline and on‐ramp CAVs while accounting for the movement of HDVs. Third, Gipps car‐following model and heuristic control are combined to represent the HDVs’ merging maneuvers. The proposed algorithm simulates and performs merging maneuvers at a typical two‐lane freeway merging area and verifies it in various scenarios considering demand level, demand splits and CAV Penetration Rate (PR). The simulation results show that the proposed algorithm can effectively facilitate merging operations, reduce the Total Travel Time (TTT), and increase the Average Travel Speeds (ATS) compared to other merging algorithms. Specifically, compared to the case of using only cooperative merging control method, the proposed algorithm can further reduce TTT by 25.5% and increase ATS by 33.38%. When PR gradually increased, the control performance of the proposed algorithm can be further improved.

show abstract

Section: Simulation Resultsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Macroscopic Traffic Flow Control At Freeway Merging Areasmentioning

confidence: 99%

Section: Macroscopic Traffic Flow Control At Freeway Merging Areasmentioning

confidence: 99%

See 3 more Smart Citations

Multilane freeway merging control via trajectory optimization in a mixed traffic environment

Han

Zhang

Guo

2023

IET Intelligent Trans Sys

View full text Add to dashboard Cite

show abstract

Leveraging reinforcement learning for dynamic traffic control: A survey and challenges for field implementation

Han,

Wang,

Leclercq

2023

Communications in Transportation Research

View full text Add to dashboard Cite

A Deep Reinforcement Learning Based Ramp Metering Control Method Considering Ramp Outflow

Cheng,

Ye,

Wang

et al. 2024

IFAC-PapersOnLine

View full text Add to dashboard Cite

Integrated Traffic Control for Freeway Recurrent Bottleneck Based on Deep Reinforcement Learning

Cited by 21 publications

References 35 publications

Multilane freeway merging control via trajectory optimization in a mixed traffic environment

Multilane freeway merging control via trajectory optimization in a mixed traffic environment

Leveraging reinforcement learning for dynamic traffic control: A survey and challenges for field implementation

A Deep Reinforcement Learning Based Ramp Metering Control Method Considering Ramp Outflow

Contact Info

Product

Resources

About