On Resource Efficient and Obstacle Aware Link Selection in D2D Communications

Abstract

Device-to-device (D2D) communication has been envisioned as the solution to the bandwidth scarcity problem in the era of exponentially growing smart handheld devices. In D2D communications, two or more user equipment (UEs) are allowed to directly communicate with each other with limited or no involvement of the base station (BS). Since the number of available frequency channels is limited, one must judiciously allocate the channel resources among the demanding UEs. In cases where the direct communication link between two UEs offers poor signal quality, an idle UE may be judiciously selected to establish a relay-aided indirect communication link. To cope with the high bandwidth demands of modern applications, D2D communication using millimeter-wave (mmWave) signals has been proposed due to its improved spectral efficiency, higher data rates, and lower delays. The major challenge of using mmWave signals is that they suffer from high penetration and propagation losses and thus require short-distance obstacle-free line-of-sight (LOS) communication. The two problems, namely channel allocation and relay selection, have inherent interdependencies and thus must be jointly dealt with. To this end, in our first work of this thesis, we have tried to address the joint relay selection and channel assignment problem (JRSCAP) for D2D communications, and devised a near-optimal algorithm with polynomial time complexity. Both user mobility and the presence of static as well as dynamic obstacles can severely affect an mmWave communication link. Next, in this thesis, we have investigated the JRSCAP for mobile UEs in the presence of obstacles. After proving the hardness of this joint problem, we provide a greedy solution along with its approximation bound. For an energy-efficient green communication network, one must jointly allocate the frequency channel to requesting users as well as control their transmit power. As mentioned, the presence of obstacles can break an mmWave communication link, which may require a retransmission and contribute to wasteful energy consumption. While static obstacles are easier to avoid, dynamic obstacles pose the main hurdle, as they move independently outside the purview of the BS. Here, we have proposed a reinforcement learning (RL) framework for the joint power and channel allocation problem (JPCAP) for maximizing energy efficiency in the presence of dynamic obstacles. Information about dynamic obstacles can also be learned from link failures. To obtain a complete knowledge about the whole service area, sometimes we may be required to nonoptimally allocate resources so that all requesting links get an equal chance of activation. Although such non-optimal allocations are undesirable, they help in acquiring information about all the links uniformly. This brings us to the infamous exploration–exploitation dilemma. To this end, we have proposed a systematic way of inducing non-optimality in JPCAP. Given the hardness of this problem, we have devised a greedy solution and shown its effectiveness. In many modern applications, such as video streaming, the same data packets may need to be delivered to a group of users. Multicasting these packets has a clear advantage over repeated unicasts. Due to the dynamic nature of wireless communication links, establishing a stable multicast communication route is a challenging task, especially in the presence of dynamic obstacles. We address the multicast link selection problem (MLSP) as our final work in this thesis and present an optimal algorithm for stable link selection in the presence of dynamic obstacles. For all of our work in this thesis, we have performed extensive simulations and shown that our proposed solutions outperform existing state-of-the-art approaches.