Q-Learning Based and Energy-Aware Multipath Congestion Control in Mobile Wireless Network

Along with the development of mobile wireless communication technologies, many devices are equipped with more than on network interfaces (4G/5G, Wi-Fi, Bluetooth, etc.). To aggregate the idle bandwidth of different network interfaces, Multipath Transmission Control Protocols (MPTCP) are standardized by the Internet Engineering Task Force (IETF). MPTCP can establish sub-flows through different network interface in one connection and improve the transmission efficiency by transmitting data concurrently. However, there are still two problem for MPTCP to work in the mobile wireless network: (1) Unawareness to the network changes; (2) No consideration of energy consumption. To address these two issues, we propose the Q-Learning based and Energy-aware Multipath Congestion Control (QE-MCC) scheme in this paper. Firstly, the stability and trend parameters are introduced to formulate the system state. Then, an energy-aware transmission utility model is presented to evaluate the effects of congestion control. Finally, the Q-learning based congestion control algorithms are designed to improve transmission efficiency. The simulation results shows that QE-MCC performs better on throughput, delay and, energy consumption compared with standard and similar solutions.