Semi-Blind Multipath Channel Estimation and Precoding Design in AF Two-Way Relay Networks

We propose a semi-blind channel estimation for two-way relay networks (TWRNs) where multiple relays employ amplify-and-forward (AF) protocol and the channel is frequency-selectively faded. The challenges of the channel estimation are resulted from inter-symbol interference (ISI), self-interference in two-way systems and difficulty to estimate each relay link separately. To deal with ISI and to obtain the effective channel impulse response (CIR) of each relay link individually, cyclic prefix is inserted in source messages and we propose a low-complexity and scalable design of precoding matrices at relays. Under the framework, effective CIR of each relay link is estimated through the second-order statistics of the received signals, while ambiguity resulted from the channel matrices of direct links between users are resolved by a small number of training blocks. It shows through simulations that NMSE of the proposed scheme exists error floor at high SNR, which can be mitigated by averaging over more blocks of received signals especially when the channel varies slowly.