Stop criteria for retransmission termination in soft-combining algorithms

Abstract

Soft-combining algorithms use retransmissions of the same codeword to improve the reliability of communication over very noisy channels. In this paper, soft-outputs from a maximum a posteriori (MAP) decoder are used as a priori information for decoding of retransmitted codewords. As all received words may not need the same number of retransmissions to achieve satisfactory reliability, a stop criterion to terminate retransmissions needs to be identified. As a first and very simple stop criterion, we propose an algorithm which uses the sign of the soft-output at the MAP decoder. The performance obtained with this stop criterion is compared with the one assuming a genius observer, which identifies otherwise undetectable errors. Since this technique needs always a particular number of initial retransmissions, we exploit cross-entropy between subsequent retransmissions as a more advanced but still simple stop criterion. Simulation results show that significant performance improvement can be gained with soft-combining techniques compared to simple hard or soft decision decoding. It also shows that the examined stop criteria perform very close to the optimistic case of a genius observer.