Channel Modeling And Its Applications In Error Estimating Coding

Channel modeling is a fundamental technology in wireless communications. It trys to employ model-based approachs to dig and figure out inherent laws behind the complex and stochastic errors in wireless data transmission, and empresses these laws qualitatively or quantitatively using the model structure and parameters. This will provide basis to the design and performance evaluation of upper-layer protocols.This dissertation uses the recorded traces in the real wireless network environment to build channel models and then studys the statistical laws of transmission errors in wireless channels using the proposed models. Finally, these laws are used to boost the accuracy of error estimating coding. The main contributions are summarized as follows.(1)Modeling of heavy-tailed wireless channelsThe heavy-tailed behavior of wireless channels fits with observations of the statistical laws of transmission errors and it plays an impact on the performance of upper-layer protocols. Therefore, it is of great theoretical and practical significance to build an accurate heavy-tailed channel model. The existing heavy-tailed channel model is affected by the complex model structure, high run-time cost, complex parameterization and so on.This dissertation starts with the run length and burst length complementary cumulative distribution function(CCDF), and proposes a novel heavy-tailed channel model based on an independent but not identically distributed(inid) stochastic process. inid has the advantage of simple and flexible model structure, low run-time cost(a Bernoulli’s test). It contains only two requested parameters and is easily parameterized.(2)Study of generic error model for wireless channelsWireless channels are intrinsically time-varying and unpredictable, which makes recorded traces to be of diversified modeling characteristics. The time-varying channels and different traces bring difficulties to channel modeling. This dissertation defines external run, internal run, error process and proposes a novel trace-representing method. Based on the method, an accurate and generic error model(GEM) is presented. GEM has the advantage of simple model structure and high applicability. All parameters have definite physical meanings and the modeling process is independent to characteristics of recorded traces. Experimental results show that GEM can be used to describe different kinds of recorded traces.(3)Study of the model-driven error estimating coding algorithmError estimating coding a coding-based packet bit error rate(BER) estimation method.The core ideas are that the transmitter sends a small error estimating codeword along with each packet that will allow the number of bits in the packet corrupted during the wireless transmission to be inferred at the receiver. The BER of the packet is closely related to error characteristics in wireless channels, thus it is of great theoretical and practical significance to consider such characteristics during the design of error estimating coding algorithms. Combining with the model-based error characteristics analysis after channel decoding, a model-driven error estimating coding algorithm(MEEC) is proposed in this dissertation. Experimental results show that MEEC can overcome bursty and short-term memory and provide higher accuracy with lower redundancy.(4)Study of model-based smart error estimating codingThe traditional philosophy behind existing error estimating coding algorithms is that a transmitter computes and encodes error estimating codewords based on the data bits before channel coding and the receiver estimates BER of a packet based on the data bits after channel decoding. We refer to it as post-decoding error estimating coding. This dissertation first proposes pre-decoding error estimating coding, namely the transmitter computes and encodes the error estimating codewords based on the data bits after channel coding and before modulation; the receiver estimates BER of a packet based on the data bits after demodulation and before channel decoding. Since the channel decoding is very time-consuming, the proposed pre-decoding error estimating coding can realize better real-time character. This dissertation employs the channel model to figure out error characteristics before channel decoding, and combined with the well-defined physical symbol error structure of M-QAM, presents smart error estimating coding(Smart-EEC).Experimental results show that Smart-EEC helps EEC and MEEC achieve a better tradeoff between the space redundancy and estimating accuracy.In summary, this dissertation investigates channel modeling, model-based understanding of error characteristics and its applications in error estimating coding. Our works enrich method of channel modeling, promote the integratioin of channel modeling and error estimating coding. The dissertation is of theory and application interest.