Performance Analysis Of Mode Selection In Device-to-Device Communications Underlaying Cellular Networks

Device-to-device (D2D) communications commonly refer to technologies that enable devices to communicate directly with each other in a single hop or along routes without the need of the base stations. The so-called device includes human and machine, which refers to Human-to-Human (H2H) via mobile phone and Machine-to-Machine (M2M) via communication without human involvement, respectively. D2D function in cellular networks is envisioned to provide better controllable interference in the licensed bands, longer distance communications and quicker connections than traditional D2D technologies (e.g. Bluetooth and Wifi-Direct).In D2D communications, mode selection is one of the key research problems, which is closely connected with the power control algorithms and resource allocation schemes. In this paper, both static and dynamic selections of different resource allocation schemes by setting different resource allocation restrictions are studied. Queuing model for D2D communications under cellular networks is developed when a mode slection algorithm is chosen. End-to-end average throughput, average packet delay and average dropping probability are analyzed and compared to decide the optimal mode slection algorithm considering different network topology and different network loads.The introduction of D2D function may bring additional interference into cellular networks, i.e, interference between multiple D2D users, interference between a cellular user and a D2D user and interference between a cellular user and multiple D2D users. Existing work on mode selection in D2D communications usually assumes that the D2D and cellular users are saturated with infinite backlogs, where all links transmit packets in every time slot and interference always exists between links using the same resource. In this paper, we study the performance of D2D communications underlaying cellular networks based on the dynamic interference model, where finite amount of packets arrive to the queues at every time slot. Thus, the interference exists only when the links reusing the resource transmit packets simultaneously.D2D devices can communicate with each other through D2D links or cellular links. When utilizing D2D links, the number of arrival packets in a time slot follows the Poisson distributed packet arrival process. When utilizing cellular links, the arrival rate of the first hop follows the Possion distribution, while the second hop is related with the departure process of the first hop.We construct Discrete Time Markov Chains (DTMC)s underlying the queuing models where every state is a tuple consisting of the queue states and the channel states of the links that may be scheduled to serve the queues. Due to the large state space and computational complexity of the exact model, we use model decomposition and iteration technique in stochastic Petri nets to decompose the exact queuing models into submodels with inter-correlated service rates. We compute the approximate stationary distributions of the DTMCs underlying the decomposed submodels, which enable us to eliminate the correlation between submodels. Fix point iteration is used to realize cyclic iteration.