Efficient Utilization Method of Motorway Lanes Based on YOLO-LSTM Model

TL;DR

This paper introduces a model combining YOLO and LSTM to manage motorway congestion by deciding when to open emergency lanes.

Contribution

A novel ML-YOLO model with DeepSORT and LSTM-Dropout for real-time traffic congestion management and emergency lane control.

Findings

Emergency lanes should be opened when the TPI exceeds 0.17 and closed when below 0.17.

The model effectively reduces congestion and improves motorway traffic efficiency.

The LSTM model's predictions showed high accuracy with low relative error.

Abstract

With the development of cities, traffic congestion has become a common problem, which seriously affects the efficiency of motorway transport. This study proposed an improved ML-YOLO video data extraction model based on You Only Look Once (YOLOv8n) combined with the Deep Simple Online and real-time tracking (DeepSORT) algorithm, to classify the obtained Traffic Performance Index (TPI) into different congestion levels by extracting traffic flow parameters in real-time and combining with the K-means clustering algorithm. The Long Short-Term Memory Dropout (LSTM-Dropout) model and the emergency lane opening model were used to implement the road congestion warning successfully. The practicality and stability of the model were also verified by calculating the relative error between the predicted traffic flow parameters and the extracted parameters through the LSTM time series model. According to the model results, emergency lanes are opened when the motorway traffic TPI exceeds 0.17 and closed when below 0.17. This study provided a reasonable theoretical basis for motorway traffic managers to decide whether or not to open the emergency lane, effectively relieved motorway road congestion, improved efficiency of road traffic, and had important practical value and significance in reality.

Figures