The Research On Confidence Probability And Efficient Algorithms In Wireless Communication Simulations

In recent years, there are great progresses in wireless communications,accordingly the system frame becomes very complicated. It is hard to achievesystem performance such as bit error rate (BER) only by mathematic analysis,so computer simulation acts a very important role. This thesis focuses on theBER estimation in link-lever simulations in wireless communications. All thediscussions are about two aspects. One researches on the numericrelationship between the confidence probability (CP) of BER estimation andthe size of simulation samples, the other on efficient simulation algorithms,which designed to reduce simulation samples under given CP.In Chapter 2, discussions spread around MC methods. Based on thecorrelation between simulation samples, MC are classified as traditional MC(TMC) and general MC (GMC). Firstly, TMC method and its conclusions areintroduced. Secondly, condition of Stable Finite Adjacent Correlation (SFAC)is proposed. GMC estimation that follows SFAC is under consideration andsome conclusions have been drawn out. Thirdly, MC methods are applied tolink-lever simulations in wireless communications. It is pointed out thatalmost all the simulation results follow TMC estimation or GMC estimationthat follows SFAC. Lastly, a universal evaluation algorithm based on SFACis proposed, which can evaluate the CP of BER estimation in general wirelesscommunication systems. In Chapter 3, the CP of BER estimation under different wirelesschannels is investigated. The channels include additive white Gaussian noise(AWGN) channel, single input single output (SISO) fast fading channel,SISO quasi-fast fading channel, SISO flat slow fading channel, SISOmultipath slow fading channel, multiple input multiple output (MIMO) fastfading channel, and MIMO flat slow fading channel. And the research underSISO multipath slow fading channel is the key part of this chapter.In Chapter 4, discussions spread around Important Sampling (IS)methods. IS methods are proposed to reduce the simulation samples byrevising the sampling density function at high signal to noise ratio (SNR).Firstly, based on the correlation between the simulation samples, IS can beclassified as traditional IS (TIS) and general IS (GIS). Firstly, TIS and GISare introduced respectively. Secondly, two universal adaptive IS algorithmsare proposed. Lastly, IS simulations are investigated under different channelsincluding AWGN channel, SISO fading channel, and MIMO fading channel.In Chapter 5, four efficient simulation algorithms are proposed. All thesealgorithms are universal and can be applied to many situations or combinedtogether. The first is Quasi-Analytical simulation algorithms, which canreduce the simulation samples by combining the mathematic analysis andcomputer simulations. The second is Adaptive Sample Size Self-Decisionsimulation algorithm, which can reduce total simulation samples by savingthe redundant ones. The third is Error Classification simulation algorithm, which can reduce total simulation samples by optimizing the sub-simulationsamples for different error kinds. The fourth is Control Variates simulationalgorithm, which can reduce the estimator's variance by eliminatingrandomicity.In Chapter 6, all the research results and ideas made in this thesis aresummarized together. Based on this thesis some prospective areas aresuggested for future research.