Pricing QoE With Reinforcement Learning For Intelligent Wireless Multimedia Communications

With the unification of modern wireless services of Mixed Reality and Tactile Internet, providing Quality of Experience (QoE) with ultra low latency become critical challenges in edge resource allocation. In this paper we propose a reinforcement learning-based economic pricing model for wireless multimedia QoE, leveraging economic theories and machine intelligence. In the proposed approach, the QoE pricing model considers the User Equipment (UE)'s perceived QoE, the amount of purchased data, the wireless channel conditions, and the user's subjective multimedia content preference. In addition, the QoE gain of UE, cost of three entities in wireless networks Content Provider (CP), Wireless Carrier (WC), and UE, are integrated in the economic concept of social utility. The social utility would be affected by all system factors such as unit data price, multimedia quality requirement, and wireless channel conditions. The proposed reinforcement learning method improves the social utility performance by maximizing the accumulated utility through obtaining the optimal factors set up. At last, through numerical simulations we show the impacts of different system parameters on UE's QoE gain and the improvement of social utility performance by using the proposed reinforcement learning approach.