AI-Enabled Traffic Control Prioritization in Software-Defined IoT Networks for Smart Agriculture | JoVE Visualize

Area of Science:

Computer Science
Network Engineering
Machine Learning

Background:

Smart agricultural systems utilize sensor networks and IoT devices to gather environmental data for improved farming efficiency.
Traditional data processing methods struggle with the complexity and dynamism of data in smart agricultural systems.
Optimizing traffic routing in Software-Defined Networks (SDN) is crucial for efficient data management and resource allocation.

Purpose of the Study:

To propose a novel framework integrating Machine Learning (ML) and Reinforcement Learning (RL) for optimizing traffic routing in SDN.
To classify data traffic into emergency, normal, and on-demand categories using various ML models.
To enhance network performance by reducing latency and improving resource allocation, especially for time-sensitive traffic.

Main Methods:

Employed ML models including Logistic Regression (LR), Random Forest (RF), k-nearest Neighbours (KNN), Support Vector Machines (SVM), Naive Bayes (NB), and Decision Trees (DT) for traffic classification.
Utilized the Q-learning (QL) algorithm, a basic RL version, within the SDN paradigm for optimizing routing based on classified traffic.
Evaluated the performance of ML models, noting superior accuracy from Random Forest (RF) and Decision Trees (DT).

Main Results:

The proposed ML and RL integrated framework effectively optimizes traffic routing in SDN environments.
Random Forest (RF) and Decision Trees (DT) demonstrated higher accuracy in classifying traffic data compared to other ML models.
The integration significantly improves resource allocation, reduces network latency, and prioritizes the delivery of emergency traffic.

Conclusions:

The combined ML-based data classification and QL algorithm offer a robust solution for dynamic traffic routing in SDN.
SDN's adaptability allows routing algorithms to adjust to real-time network conditions and traffic characteristics.
This approach is vital for enhancing the efficiency and responsiveness of smart agricultural systems and other data-intensive applications.