Research On IDMA Technology And Power Optimization

One of the greatest challenges of wireless communication is the efficient utilization of the scarce resource which is the wireless spectrum. This problem has become more acute due the continuous increase in the number of mobile users all over the globe. Knowing that the wireless spectrum is shared by all users and therefore interfering with each other, a few methods have been utilized so far for allocating non-overlapping resources (either in time, frequency or code) to each individual user. The main focus in this thesis is the study of a method called Interleave-Division Multiple-Access (IDMA), which consists of allocating all the available resource to the users at the same time. In this multiple-access scheme the users are only required to have randomly generated chip-level interleavers. In this thesis, a method is proposed for optimally allocating power to users such as to increase the overall system spectral efficiency and some methods are proposed for reducing the implementation complexity of IDMA in eventual wireless communication systems.IDMA has the merit to exceed the spectral efficiency of classic spread spectrum systems such as DS-CDMA and to further increase the efficiency with proper power optimization. A method based on evolutionary algorithm called Differential Evolution (DE) is proposed to solve the power optimization problem of IDMA with a proper constraint handling procedure. This method is particularly appealing since classic optimization algorithms are not applicable due to the complexity of the problem which is iterative, non-convex and includes functions without explicit expression. Compared to the Linear Programming (LP) based algorithm, the proposed method need not to partition users in equal power groups or quantize powers, but nevertheless achieves better performances and is applicable for solving the receiver and transmitter power problems, and the asymptotic power optimization case as well. Knowing that DE belongs to the family of genetic algorithms, it suffers from slow convergence with high-dimensionality problems. Consequently, in order to apply DE to the optimization problem at stake a method to accelerate the convergence speed called gene sorting is proposed. The main idea is to arrange the coordinates of individuals in the population during iterations so as to converge to the optimum faster. In order to prove its effectiveness, the concept is studied on a benchmark set of 18 functions, three non-adaptive and three self-adaptive variants of DE, respectively.Another aspect of the asymptotic optimization problem considered is to find an analytic expression of the solution. However, in the state-of-art, equal receiver power profile also called equal power allocation (EPA) is the only solution of this kind found so far. In this thesis, a constraint is formulated and has to be met for EPA to be feasible. It is also shown that when this constraint is met, then the transmitter and receiver power optimization problems are equivalent.Moreover, a simple interleaver generation method based on linear congruential generator (LCG) is proposed. This method enables the users to easily generate their own interleavers and consequently avoiding the waste of bandwidth due to interleaver pattern's transfer. It consists of choosing parameters such that the LCG period is longer than the interleaver size and out-of-range numbers are discarded therefore creating more randomness in the sequence. One interleaver called mother interleaver or master interleaver can be generated this way and other obtained by cyclically shifting it. Another equivalent generation method is by using a specific seed for each interleaver. Results show that the interleavers obtained by the proposed method have close performance to the all random interleavers and is therefore more appealing in an implementation point of view.In the same prospective of simplified implementation, the application of variable spreading factor (VSF) in IDMA is proposed for enabling different users to meet different throughputs, and therefore achieving different QoS. Most importantly, the VSF approach is a convenient way to avoid the PAPR problem induced by layer superposition. Furthermore, in order to trade-off between performance and complexity/latency the Group-Wise Successive Interference Cancellation (GSIC) method is proposed to combine the fast convergence performance of Successive Interference Cancellation (SIC) and the low latency of Parallel Interference Cancellation (PIC).