Optimal Training Sequence Design for Bi-Directional Relay Networks

In this paper, the problem of training sequence design in a bi-directional relay network is investigated. The two nodes Txl and Tx2 communicate with each other via a relay (RL). The RL receives the signals that are transmitted simultaneously from both nodes and then broadcasts the superposition back to the two nodes. By doing so, bi-directional relaying reduces half the channel resources required by the conventional relaying. In frequency-selective fading channels, single carrier cyclic prefic (SCCP) is efficient in combatting intersymbol interference (ISI). Hence, in this paper, SCCP is deployed in the transmission from the two nodes and the RL as well. We are interested in the channel estimation problem in this scenario. We propose a design of training sequences from two nodes to minimize the mean square error (MSE) of the channel estimation according to zero forcing (ZF) criterion. We do not separate the links between each node to RL but corporate them into composite channel. Simulation results show that our design performance suffers only 1 dB loss in terms of bit-error-rate (BER) as compared to the case with perfect channel information case.