Optimization Design Of Mode Selection Mechanism For D2D Communication Underlaying Cellular Networks

With the rapid development and popularization of mobile terminals, as well as the swift increase of data services, the tranditional wireless cellular communication faces a great challenge. Introducing D2D (Device-to-Device) technology to cellular network can increase spectrum efficiency and capacity of system. Thus, D2D communication plays a more and more important role in the future wireless communication study. However, the user interference scenario of cellular network with D2D technology is thoroughly different, the new design of optimal resources control is an important research direction, and the mode selection is one of the key technologies.At present, most related works usually assume that the D2D and cellular users are saturated with infinite backlogs, but this paper focuses on the study of mode selection mechanism for D2D communication underlaying cellular networks based on the packet-level dynamic traffic arrival. Introducing D2D technology to cellular network adopting orthogonal frequency division multiplexing access technology, the mode selection and resource allocation are combined together, the optimization problem of this paper is minimizing the system delay under the dropping probability constraint over all users. Using queuing theory and Markov decision process as theoretical tool, the queuing model is developed, and the stochastic optimization model of mode selection mechanism——Markov decision process model is built, the proposed constraint optimization problem is formulated as an infinite average reward constraint Markov decision process model.Based on the infinite average reward constraint Markov decision process model, to solve the model, firstly, the Lagrangian multiplier method is utilized to turn the constraint Markov decision process (CMDP) model into an unconstraint Markov Decision Process (MDP) model. Since the formulated MDP model is faced with the "curse of dimensionality" problem, in this paper, the equivalent Bellman’s equation is introduced to reduce the system state space; the linear approximation method is introduced to further reduce the number of value functions, and a subchannel bidding mechanism is proposed to obtain the optimal control policy; the online stochastic learning algorithm is introduced to establish the value functions and Lagrangian multipliers (LMs) by means of online update, and the convergence proof and approximate optimal proof are presented. Based on the equivalent Bellman’s equation, linear approximation method and online stochastic learning algorithm above, this paper proposes a distributed subchannel allocation algorithm to realize the mode selection and resource allocation with optimal system delay. The simulation platform is set up for performance verification and comparative analysis of proposed algorithm, the simulation results show that the performance of proposed algorithm is close to that of the offline value iteration algorithm and better than that of two other reference algorithms, which verifies the approximate optimal performance of the proposed distributed subchannel allocation algorithm.