ML performance analysis of the decode-and-forward protocol in cooperative diversity networks

Abstract

We analyze the maximum-likelihood (ML) performance of the decode-and-forward protocol in a cooperative diversity network which consists of a source, a relay, and a destination with a direct path signal, but which is not equipped with cyclicredundancy- check (CRC) codes. In this system, due to a symbol error at the relay, the ML receiver at the destination needs to consider all the possible symbol detection scenarios at the relay as well as at the destination. Therefore, the ML detection metric is given by a linear combination of exponential functions, which prevents the use of the classical minimum Euclidean distance rule. Adopting the max-log approximation, we approximate the ML detection rule which makes the ML performance analysis tractable. In order to facilitate the derivation of decision regions, we simplify the ML detection rule in the two-dimensional real space such that two metric values of two adjacent constellation points are sequentially compared. Then we obtain decision regions in a form without union and intersection. Finally, based on the decision regions, we derive a very accurate closedform BER approximation for M-pulse amplitude modulation (PAM) and M-quadrature amplitude modulation (QAM). The obtained BER expression can serve as the error performance upper-bound of the decode-and-forward protocol in cooperative diversity networks.