A Research On The Cooperation Technologies Between Transmitters And Receivers Under Time Varying Cooperative Diversity Channels

Cooperative diversity scheme has the potential to be the key communication technology in the future, because of its high reliability, high transmission rates and ubiquitous coverage. The premise of exploring these advantages is obtaining accurate channel state information (CSI) through channel estimation. Therefore, the design of channel estimation is crucial for cooperative diversity system. This thesis focuses on the challenges faced by channel estimation of time varying cooperative diversity systems and proposes new optimized designs for training parameters, signal detection and relay strategies.To describe the achievable performance of realistic time varying cooperative diversity systems, this thesis proposes the achievable rate expressions for both Decode-and-Forward (DF) and Amplify-and-Forward (AF), after considering the channel estimation error and resources occupied by training symbols. Based on this, we analyze and optimize two main training designs: Pilot symbol assisted modulation (PSAM) training and Superimposed training (ST).For the time multiplexed PSAM training, this thesis proposes the optimization strategies for the two training parameters: bandwidth efficiency factors and power allocation factors, by maximizing the achievable rate obtained previously. The optimization results are given by closed expressions for both DF and AF. Simulations verify that our proposed optimization strategy can achieve considerable performance gains and alsohas negligible loss compared to the optimal designs.To resolve the problems of noise propagation and Doppler frequency cumulation faced by time varying AF systems, this thesis proposes a new relaying strategy: Equalization-and-Forward (EAF). EAF carries out channel estimation and equalization at the relay node and is more suitable for time varying environment. We utilize both PSAM and ST in EAF and compare it with the ordinary AF through both theoretical analysis and realistic simulations. It is verified that EAF can obtain a gain of 4~5dB by small complexity increase at the relay node.For superimposed training, this thesis points out the problem that the ordinary signal detection regards the Data Dependent (DD) sequences as pure noise, and by exploring the useful information embedded in DD sequences, we propose two improved detection algorithms: detection based on joint linear equalzation and joint maximum likelihood sequence detection (MLSD). Through the simulations, our proposed joint MLSD algorithm is shown to improve the performance by 2~3dB. This thesis also compares the achievable rates and BER performance of PSAM and Superimposed training under time varying channels and gives suggestions on the training selection.