Q-Learning Based D2D Offloading for Statistical QoS Provisioning Over 5G Multimedia Edge-Computing Wireless Networks
Jingqing Wang, Xi Zhang · 2019
With the increasing demand for the wireless multimedia services over 5G multimedia edge-computing wireless networks, the statistical quality-of-service (QoS) provisioning has been proven to be able to effectively guarantee the multimedia data transmissions over time-varying wireless channels. On the other hand, the collaborative device-to-device (D2D) based offloading scheme, where the mobile users share the popular multimedia files within a D2D communication group instead of downloading from the remote backhaul networks, has been proposed as one of the 5G promising candidate techniques to meet the statistical delaybounded QoS requirements as well as address the data explosion problem introduced by 5G multimedia edge-computing wireless networks. However, due to the dynamically changing environment and various demands for each mobile user, it is challenging to design an efficient offloading algorithm for choosing the optimal D2D offloading strategy. To overcome the aforementioned problems, in this paper we propose the machine-learning based framework for the collaborative D2D offloading scheme through repeatedly interacting with the networking environment while satisfying the statistical delay-bounded QoS constraints. In particular, we derive and analyze the collaborative D2D communication model and the interference based constraints. Under the statistical delaybounded QoS requirements, we formulate and solve the effectivecapacity optimization problem for the collaborative D2D offloading scheme by using Q-learning technique over 5G multimedia edgecomputing wireless networks. Also conducted is a set of simulations which evaluate the system performance and show that our proposed collaborative D2D offloading scheme outperforms the other existing schemes under the statistical delay-bounded QoS constraints over 5G multimedia edge-computing wireless networks.