Methodology for the Identification of Vehicle Congestion Based on Dynamic Clustering

Reyes, Gary; Tolozano-Benites, Roberto; Lanzarini, Laura; Estrebou, César; Bariviera, Aurelio F.; Barzola-Monteses, Julio

doi:10.3390/su152416575

Cited by 1 publication

(1 citation statement)

References 41 publications

(45 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The input data set can be coded in such a way as to include external factors (for example, weather conditions) [13,14]. However, increasing the complexity of the input data set will increase both the training time and the prediction time in the testing phase.…”

Section: Introduction and Related Workmentioning

confidence: 99%

Graph-Based Neural Networks’ Framework Using Microcontrollers for Energy-Efficient Traffic Forecasting

Zoican,

Galatchi

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

This paper illustrates a general framework in which a neural network application can be easily integrated and proposes a traffic forecasting approach that uses neural networks based on graphs. Neural networks based on graphs have the advantage of capturing spatial–temporal characteristics that cannot be captured by other types of neural networks. This is due to entries that are graphs that, by their nature, include, besides a certain topology (the spatial characteristic), connections between nodes that model the costs (traffic load, speed, and road length) of the roads between nodes that can vary over time (the temporal characteristic). As a result, a prediction in a node influences the prediction from adjacent nodes, and, globally, the prediction has more precision. On the other hand, an adequate neural network leads to a good prediction, but its complexity can be higher. A recurrent neural network like LSTM is suitable for making predictions. A reduction in complexity can be achieved by choosing a relatively small number (usually determined by experiments) of hidden levels. The use of graphs as inputs to the neural network and the choice of a recurrent neural network combined lead to good accuracy in traffic prediction with a low enough implementation effort that it can be accomplished on microcontrollers with relatively limited resources. The proposed method minimizes the communication network (between vehicles and database servers) load and represents a reasonable trade-off between the communication network load and forecasting accuracy. Traffic prediction leads to less-congested routes and, therefore, to a reduction in energy consumption. The traffic is forecasted using an LSTM neural network with a regression layer. The inputs of the neural network are sequences—obtained from a graph that represents the road network—at specific moments in time that are read from traffic sensors or the outputs of the neural network (forecasting sequences). The input sequences can be filtered to improve the forecasting accuracy. This general framework is based on the Contiki IoT operating system, which ensures support for wireless communication and the efficient implementation of processes in a resource-constrained system, and it is particularized to implement a graph neural network. Two cases are studied: one case in which the traffic sensors are periodically read and another case in which the traffic sensors are read when their values’ changes are detected. A comparison between the cases is made, and the influence of filtering is evaluated. The obtained accuracy is very good and is very close to the accuracy obtained in an infinite precision simulation, the computation time is low enough, and the system can work in real time.

show abstract

Section: Introduction and Related Workmentioning

confidence: 99%