Adaptive Modulation and Coding in Wireless Communications with Feedback
Doctoral thesis, 2002
We study link adaptation techniques which are suitable for the future packet-based wireless communication systems, where in particular, flat Rayleigh fading channels are assumed.
Firstly, an adaptive coding scheme based on Rate Compatible Convolutional (RCC) codes is proposed. This scheme provides a reliable transmission with high throughput without access to the Channel State Information (CSI), through Incremental Redundancy (IR) incorporated with Automatic Repeat reQuest (ARQ). The transmission starts with code rate 1 and continues with lower rate codes on retransmission requests. The proposed scheme is further improved by also employing high level modulations. On retransmissions, the constellation size as well as the coding rate are reduced. We also investigate methods for designing an adaptive modulation and coding scheme with a limited number of transmissions. The competitive scheme is subjected to maximize the throughput for all the users over the coverage area.
In the second approach, channel prediction is considered in the system. An optimum unbiased predictor in the mean square sense is used to predict the channel SNR based on the past noisy estimates at the receiver. The predicted CSI is sent to the transmitter via the feedback channel. The optimum design of an adaptive modulation scheme based on uncoded M-ary Quadrature Amplitude Modulation (M-QAM) is investigated here. The transmitter adjusts the transmission rate and possibly also the power, based on the predicted Signal-to-Noise Ratio (SNR) to maximize the spectral efficiency while satisfying the BER and average transmitted power constraints. The statistical information about the prediction error is taken into account in the design. Finally, a more complete system solution is studied, where an adaptive packet transmission scheme at the link layer in combination with a time-slot scheduler supplied by channel predictions is considered. Different traffic classes with different target BER and priorities are assumed and the impact of imperfect prediction on the link layer performance is investigated.