UBC Theses and Dissertations
Performance evaluation of adaptive hybrid ARQ systems Bakhtiyari, Sattar E.
In the analysis of ARQ schemes, it is typically assumed that the feedback channel is noiseless and perfect synchronization can be achieved. These assumptions may not be valid for the situations where the channel error rate is high especially with memory ARQ schemes. In such cases, a more realistic evaluation of the performance is required to provide accurate results for the schemes in question. A general investigation of ARQ schemes with and without memory, taking into account the sensitivity of frame header and return channel errors, is presented and the resulting throughput expressions are obtained. Knowing the influence of these parameters on the system performance, we propose efficient techniques to reduce their effects on the performance degradation. These techniques range from employing more powerful error correction codes to transmitting a few copies of the same sequence for the header and acknowledgment message consecutively and combining the noisy sequences at the receiver. Even though ARQ schemes offer appreciable throughput improvement in AWGN, their performance suffers greatly in the Rayleigh fading environment. In such a scenario, the optimum solution is to adapt the rate of error correction code to the prevailing channel conditions. An adaptive Type-I Hybrid ARQ (HARQ) scheme, in which the transmitter selects the code rate for each message based on the assessed channel conditions, is proposed and investigated. A model for evaluating the throughput of the scheme is presented, and it is illustrated that an accurate assessment of the channel conditions is crucial in the achievement of high throughput. Moreover, an adaptive Type-II HARQ with Complementary Punctured Convolutional Codes (CPC) is presented. This protocol stores the corrupted copies of a packet encoded with CPC codes at the receiver and combines them to improve the reliability of the system and reduce the number of unnecessary retransmissions.
Item Citations and Data