| dc.description.abstract | In this thesis, we consider two-way relaying in a wireless communication system with two transceivers and one relay node. In such a model, two scenarios are taken into account: a system with one energy harvesting (EH) relay node and a system with two EH transceivers and one EH relay node. Under these two assumptions, we study the optimization problem of harvesting energy allocation which maximizes the total data throughput of a wireless network under some restrictions at the EH node(s). In the first scenario, only the relay node is powered up by the harvested energy and the other two transceiver nodes are powered up by conventional power supply. We first find that the data transmission rate of the system monotonically increases and is strictly concave in the power of the relay. As a result, we can obtain an optimal power control policy of the relay for maximizing the total throughput. This optimization problem can be solved not only by conventional optimization solvers, but also by a method called breaking-rope. Finally, we propose a novel algorithm to solve the optimization problem based on the breaking-rope. For the case of system with three EH nodes, the problem is more complicated since multiple optimization parameters, which are the transmitting power of three EH nodes, should be taken into account. Because the objective function is not jointly concave in all three parameters, it is difficult to find a feasible solution. Hence, we have to simplify the objective function to overcome the difficulty in solving the original problem. By relaxing the problem, we aim to find the upper bound of the total throughput. It is shown that the upper bound function is concave in one parameter when other two parameters are fixed. Thus, the optimal solution can be obtained by using the iterative Alternate Convex Search method. The performance of the proposed schemes are evaluated by computer simulations. The breaking-rope method is capable of acquiring the optimum policy in the one EH node system. For the three EH node system, we show that our solution based on the Alternate Convex Search method performs closely to the best-case and outperforms the breaking-rope method. | en |
