Multi-path transmission control protocol (MPTCP) is an extension of TCP that enables the concurrent transmission of information through different network interfaces (e.g., Cellular, Wi-Fi, 802.11p, and so on) available at terminal side. It is well known that MPTCP can provide significant advantages in bandwidth aggregation and transmission stability. Unfortunately, path diversity can limit bandwidth aggregation efficiency and incur higher delays. These issues become critical when in presence of emerging mobile AR and VR applications, which are bandwidth hungry, time-sensitive and exhibit abrupt variations of the bitrate. To address these issues, we propose theReinforcementLearning-based mobile AR/VR multipath transmission with streamingPowerSpectrumDensity analysis (RL-PSD). RL-PSD analyses the Power Spectrum Density (PSD) of the AR/VR input stream to extract its features. Then, both the input stream and network features are considered to model the MPTCP congestion control as an reinforcement learning process. Finally, a two-stage reinforcement algorithm is proposed to optimize transmission performance. RL-PSD has been tested in both single-terminal and multi-terminal scenarios: results show that it outperforms the other advanced solutions conceived to support the multipath transmission of AR/VR streams.

Multipath TCP (MPTCP) can take advantage of multiple paths to transmit data and has been deeply optimized by many researchers. However, most researchers only devote themselves to improve the transmission performance, neglecting the price cost which is another factor that users are concerned about. This paper proposes a novel stochastic optimal scheduler for MPTCP (SOS-MPTCP) that utilizes Lyapunov optimization technique in software defined wireless network (SDWN). SOS-MPTCP analyzes and solves the trade-off problem between the performance and price cost from users' perspective. Besides, with the help of centralized optimization in SDWN architecture, the controller can feed status information of each path back to mobile terminals for SOS-MPTCP to make decisions. SOS-MPTCP includes three control decisions: 1) packets admission control; 2) packets distribution control; 3) data traffic purchasing control. SOS-MPTCP aims to maximize the throughput and minimize the price cost for users. Experiment results have proved the efficiency of trade-off optimization and the transmission system can achieve the expected stability.

Multipath TCP (MPTCP), as a promising transmission protocol, can aggregate the bandwidth of multiple paths in order to improve the transmission rate. However, due to the lack of perceiving the network status from lower layers, MPTCP cannot adaptively adjust the number of subflows, which will lead to network congestion or underutilization of network resources. Besides, plenty of packets will be out-of-order severely due to the diversity among the paths so that transmission performance degrades significantly. To address the problems mentioned above, we propose a novel QoS-driven and SDN-assisted MPTCP path selection scheme (QSMPS) for high-quality transmission service. QSMPS utilizes a scalable SDN-assisted approach to monitor and analyze network status information. Through matching service demand and the provided capacity of current network, the scheme calculates the optimal number of subflows, then distributes them to the least differential delay paths determinately. Simulation results show QSMPS outperforms the existing solutions by evaluating performance in Mininet emulator and Ryu controller.