Multipath TCP-Based IoT Communication Evaluation: From the Perspective of Multipath Management with Machine Learning

Area of Science:

• Computer Science
• Networking
• Machine Learning

Background:

• Modern Internet-of-Things (IoT) devices utilize multiple network interfaces, enabled by advancements in wireless networking.
• Multipath TCP (MPTCP) enhances data transmission throughput but faces challenges with traditional path management, leading to performance degradation.

Purpose of the Study:

• To address MPTCP path management issues by integrating machine learning.
• To propose and evaluate an adaptive path selection mechanism for MPTCP.

Main Methods:

• Introduction of an automatic learning selection path mechanism based on MPTCP (ALPS-MPTCP).
• Simulation experiments comparing four machine learning algorithms for path quality assessment.
• Evaluation based on accuracy and running time.

Main Results:

• The ALPS-MPTCP mechanism adaptively selects high-quality paths for simultaneous data transmission.
• The random forest algorithm exhibited superior performance in judging path quality compared to other tested algorithms.
• Consideration of both accuracy and computational efficiency (running time) was key in algorithm selection.

Conclusions:

• Machine learning integration significantly enhances MPTCP path management.
• The random forest algorithm is a highly effective tool for real-time MPTCP path quality assessment.
• ALPS-MPTCP offers a promising solution for optimizing data transmission in multi-interface IoT environments.