Improvements and New Techniques for Deterministic and Stochastic SIR based Power Control for 3-G Wireless CDMA Networks

本文的研究内容是利用最优控制理论和现代数值线性代数技术对现有的功率控制方案进行改进，建立和求解更具一般性和更具现实意义的功率控制公式。该方案的前两部分涉及确定性功率控制问题及其变体，第三部分是同一问题的随机版本。在所提出的研究的第一部分中，通过使用不动点迭代的加速技术和Krylov子空间迭代(目前被认为是解线性大规模代数方程组的最有效的数值方法)，对目前被文献认为是最有效的分布式约束功率控制算法(PCPC)进行了改进。方案的第二部分基于作者和他的博士生最近的研究工作，考虑了一种适用于第三代无线码分多址网络的基于SIR的最优快速闭环功率控制方案。理论上得到的格式不仅是最优的，而且在一次迭代中收敛。该方案完全遵循信道变化，并假设链路在时间上不断变化。由于该格式需要一个标量离散时间估计器(预报器)，因此它实际上在4-5次迭代中收敛。在一个码分多址系统上的仿真验证了该最优功率控制算法的有效性及其相对于相应的IS-95算法和当前版本的DCPC算法的优越性。作为未来的研究课题，首先对优化后的功率控制算法与改进的DCPC算法(基于加速算法和基于Krylov子空间的算法)进行了比较分析。其次，在优化过程中，将发射功率和信干比误差的加权和联合优化，将功率控制问题置于纳什动态博弈的框架中。将优化后的功率控制结果推广到无线网络中用户之间的冲突情况，从而建立了一个纳什动态博弈问题。本建议的第三部分涉及功率控制问题的随机表述。这是一个相当新的、广泛开放的研究问题。该基本问题将确定性最优功率控制问题推广到随机环境，得到了加性高斯白噪声下的线性功率演化方程。这个问题的表述假设链路增益或背景噪声或两者都被建模为高斯白噪声随机过程。对其各种有色背景噪声和有色链路增益噪声也提出了进一步的研究方向。此外，还将考虑使用最优控制理论得出的与状态和控制相关的噪声的更实际情况。控制变量是那些在噪声环境中调节状态变量(传输功率和SIR误差)收敛到其最佳值的变量。注意，最近在最优控制文献中已经解决了离散时间域中依赖于状态和控制的噪声最优控制问题。在该方案的最后阶段，建议研究假设链路增益按泊松随机过程变化的功率控制问题。拟议研究的影响将在改善3-G无线码分多址网络的容量、服务质量(Qos)的可靠性和用户电池寿命的耐用性方面。所有这些都将通过以最佳方式控制干扰来实现。

The proposed research is about the use of optimal control theory and modern numerical linear algebratechniques to improve existing power control schemes and to formulate and solve more general and morerealistic formulations of power control problem in 3-G wireless CDMA networks. The first two parts of theproposal deal with the deterministic power control problem and its variants, and the third part is onstochastic versions of the same problem. In the first part of the proposed research, an improvement of distributed constrained power controlalgorithm (PCPC), which is presently considered in the literature as the most efficient one, is suggestedvia the use of the acceleration techniques for fixed-point iterations and via the use Krylov subspaceiterations (presently considered as the most efficient numerical method for solving systems of linearlarge scale algebraic equations). The second part of the proposal considers an optimal fast closed-loop SIR-based power controlscheme for 3-G wireless CDMA network, based on the recent research work of the author of this proposaland his doctoral student. In addition of being optimal, the scheme obtained theoretically converges in oneiteration. This scheme follows perfectly channel variations and assumes that the link gains constantlychange in time. Due to the fact that the scheme needs a scalar discrete-time estimator (predictor), itpractically converges in 4-5 iterations. Simulations on a CDMA system demonstrate the effectiveness ofthe optimal power control algorithm and its superiority over the corresponding IS-95 algorithm and thepresent version of the DCPC algorithm. As a future research topic, a comparative analysis between theoptimized power control algorithm and the improved versions of the DCPC algorithm (accelerated andKrylov subspace based) to be obtained in the first part of this proposal, is suggested first. Secondly, sinceduring optimization, the weighted sum of the transmission power and the SIR (signal-to-interference)error, are jointly optimized, the power control problem can be put in the framework of Nash dynamicgames. It is suggested to extend the optimized power control results to capture the conflict situationamong users in a wireless network, which leads to the formulation of a Nash dynamic game problem. The third part of this proposal deals with the stochastic formulation of the power control problem. Thisis a pretty much new and widely open research problem. The basic problem formulation extends thedeterministic optimized power control problem to a stochastic environment, and leads to the linear powerevolution equation with additive Gaussian white noise. This problem formulation assumes that either linkgains or background noise or both are modeled as Gaussian white noise stochastic processes. Its variantof colored background noise and colored link gains noise are also suggested for future research. Inaddition, more realistic situations of state- and control-dependent noise using optimal control theoryresults will be considered. The control variables will be the ones that regulate the convergence process ofthe state variables (transmission powers and SIR errors) to their optimal values in a noisy environment.Note that the state- and control-dependent noise optimal control problems in discrete-time domain havebeen solved recently in the optimal control literature. In the final stage of this proposal, it is suggested tostudy the power control problem assuming that the link gains change according to Poisson stochasticprocesses. The impact of the proposed research will be in the area of improvements of the capacity of 3-Gwireless CDMA networks, reliability of quality of service (QoS) and durability of user's battery life. Allthese will be achieved by controlling interference in an optimal manner.