Proactive C-ITS Decentralized Congestion Control Using LSTM

Vehicle density and channel utilization can vary significantly over short time intervals in complex and dynamic environments such as vehicular networks. In such situations, the European ITS-G5 technology proposed Decentralized Congestion Control (DCC) techniques, such as the reactive Transmit Rate C...

Full description

Saved in:
Bibliographic Details
Main Authors: Rachedi, Abdennour, Ahmed, Toufik, Mosbah, Mohamed
Format: Conference Proceeding
Language:English
Subjects:
CBR
DCC
Heuristic algorithms
ITS-G5
LSTM
Prediction algorithms
Recurrent neural networks
Resource management
RNN
Simulation
Stability analysis
Time series analysis
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Vehicle density and channel utilization can vary significantly over short time intervals in complex and dynamic environments such as vehicular networks. In such situations, the European ITS-G5 technology proposed Decentralized Congestion Control (DCC) techniques, such as the reactive Transmit Rate Control (TRC) and the adaptive Dual α algorithms, to prevent performance degradation. This paper proposes a proactive DCC technique that uses a Recurrent Neural Network (RNN), namely the Long Short-Term Memory (LSTM), to optimize the channel performance. We specifically use smoothed Channel Busy Ratio (CBR) forecasting time series calculated using an LSTM agent and provided as inputs to DCC algorithms to achieve faster channel load convergence and, as a result, improve network stability, channel load limitation, and resource allocation fairness. According to the simulation results, the proposed proactive DCC algorithms outperform existing algorithms in terms of reduction in average channel load and faster channel convergence.
ISSN:1938-1883
DOI:10.1109/ICC45855.2022.9838640