Spectrally-efficient approaches to channel estimation for amplify-and-forward two-way relay networks
Saeed Abdallah · 2013
Relay networks constitute one of the key technologies that are being developed for use in next generation wireless systems. In relay networks, the communication between the source and the destination is aided by dedicated nodes (relays) that convey the source's message to the destination. The use of relays improves the coverage, capacity and reliability in the network. Two-way relay networks (TWRNs) have recently been proposed to support bidirectional communication and have attracted the attention of many researchers because of their high spectral efficiency. In particular, TWRNs employing the amplify-and-forward (AF) protocol are appealing because of the minimal processing requirements at the relay. Effective operation of AF TWRNs requires accurate channel state information for self-interference cancellation and coherent decoding. The majority of works on channel estimation for AF TWRNs follow the training-based approach, which requires the transmission of pilots known to both terminals. The training-based approach consumes much needed bandwidth resources, which undermines the spectral efficiency of TWRNs. Blind channel estimation avoids the costly training burden by relying only on the received data samples. Another alternative approach is semi-blind estimation, a hybrid of blind and training-based approaches. The main objective of this thesis is to investigate blind and semi-blind channel estimation for AF TWRNs as a means for achieving substantially better tradeoffs between accuracy and spectral efficiency than possible using the training-based approach. In the first part of the thesis, we consider blind channel estimation for flat-fading channel conditions. Using the deterministic maximum likelihood (DML) approach, we propose new algorithms for blind channel estimation in AF TWRNs that employ constant-modulus signalling. Assuming M-PSK modulation, we prove that the proposed estimators are consistent and approach the true channel with high probability at high SNR. Using simulations, we show that the DML estimator offers a superior tradeoff between accuracy and spectral efficiency than the pilot-based LS estimator. Still within the context of flat-fading channels, the second part of the thesis focuses on semi-blind channel estimation. We derive the exact CRB for semi-blind channel estimation in AF TWRNs that employ square QAM. The derived bound is based on the true likelihood function that incorporates the exact statistics of the transmitted data symbols. Using the new bound, we show that the training overhead can be significantly reduced by employing semi-blind estimation. To demonstrate the achievability of these gains, we derive an expectation maximization (EM)-based semi-blind algorithm that performs very closely to the derived CRB. In the last part of the thesis, we consider semi-blind channel estimation for OFDM-based TWRNs operating in frequency selective channel conditions. To assist in the estimation of the individual channels, superimposed training is adopted at the relay. Our proposed semi-blind estimation algorithm is based on the Gaussian ML approach. We design the pilot vectors of the terminals and relay to optimize estimation performance. Our simulations show that the proposed method provides significant improvements in estimation accuracy.