Quality of service enabled resource scheduling for cooperative driving provision in the internet of vehicles

Ahmad, Ishfaq; Kalsoom, Nafeesa; Khalid, Samina; Ahmed, Zaheed

doi:10.1504/ijmc.2022.125424

Cited by 1 publication

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Critical to any data analysis process, data preparation takes on added significance when working with imperfect data, such as the wrong and outlier data that results when the real value does not match the computed value by the indicator over a certain time period [13]. Longer wait times at junctions owing to inaccurate data can be frustrating for drivers and contribute to pollution from idling vehicles [14]. Congestion at junctions and improper timing of signals are only two examples of how faulty information may compromise safety.…”

Section: Introductionmentioning

confidence: 99%

A Deep Reinforcement Learning-Based RNN Model in a Traffic Control System for 5G-Envisioned Internet of Vehicles

Macherla,

Muvva,

Lella

et al. 2024

MMEP

View full text Add to dashboard Cite

In metropolitan areas, traffic jams on city streets are a major source of annoyance and financial losses. Recent advancements in data processing algorithms and the widespread availability of traffic detectors have made it possible to implement data-driven strategies for reducing traffic congestion. In order to benefit from intersection cooperation in this setting, this paper presents a distributed control strategy based on RL. In this scenario, traffic prediction software's embedding that takes into account the state of nearby junctions is used to synthesize an RL controller that controls the traffic lights. Loop detector characteristics are insufficient for precise data imputed in sophisticated traffic control systems. Most current imputation methods only use these extracted characteristics, which leads to the creation of data replicas that lack the necessary precision. The clean data are first given a statistical multi-class label, with classes ranging from C1 to Cn. Then, using a deep recurrent neural network (RNN) model, the best data model is created from the labelled spotless data and applied to the class of models in the missed-volume data. Results from simulations using TRANSYT demonstrate that the suggested strategy outperforms conventional methods in terms of waiting times and other important presentation indices.

show abstract

Section: Introductionmentioning

confidence: 99%