Fast converging auction-based resource allocation for QoE-driven wireless video streaming

Quality-of-Experience (QoE)-driven centralized resource allocation approaches for video streaming have been studied intensively but require the availability of utility information about the video content and channel conditions of all users at the central optimization entity to perform an optimal allocation. On the other hand, the practical application of game-theory-based decentralized resource allocation approaches for potentially selfish users has been limited so far. This is mainly because the varying network conditions and the delay constraints for wireless multimedia communications require low-complexity methods with fast convergence. We propose an auction-based radio resource allocation method which is shown to converge within a bounded number of iterations. The achieved allocation maximizes the average QoE over all users, while the users are maximizing their own payoff. The users pay a price for the requested resources which is defined on the utility scale. The proposed game-theory framework is compatible with cross-layer optimization approaches, as the resources are abstracted to provide an interface between the application and the lower layers. We implement the proposed resource allocation scheme in a simulated LTE uplink environment with multiple video streaming users. Experimental results confirm the derived properties and additionally show that, unlike state-of-the-art decentralized resource allocation schemes, our proposed auction is scalable, as the number of iterations to converge decreases with an increasing number of participating users.