Machine learning based path management for mobile devices over MPTCP

Recent mobile devices are equipped with multiple network interfaces such as LTE and Wi-Fi. Transport protocols that can transfer data over multiple paths, especially MPTCP (Multipath TCP), allows the devices like smartphones and tablets to exploit both interfaces concurrently. However, in real environments, wireless devices abound and network quality changes frequently. It makes network connection affect the MPTCP performance negatively. In this paper, we propose a novel path management scheme called MPTCP-ML (MPTCP based on Machine Learning) to make MPTCP troubleshoot the problem. It manages path usage among multiple connections based on decision computed by machine learning model. For accurate capturing of path quality, we utilize various quality metrics including signal strength, data rate, TCP throughput, the number of interference APs, and RTT (Round Trip Time). We have implemented MPTCP-ML in Android and conducted experiments for various and dynamic environments. The results show that MPTCP-ML outperforms generic MPTCP, especially for mobile environments.