Towards near optimal network coding for secondary users in cognitive radio networks

In cognitive radio networks (CRNs), secondary users (SUs) may employ network coding to pursue higher throughput. However, because SUs should not interfere with high-priority primary users (PUs), the available transmission time of SUs is usually uncertain, i.e., SUs do not know how long the idle state can last. Meanwhile, existing network coding strategies generally adopt a block-based transmission scheme, implying that all packets in the same block can be decoded simultaneously only with enough coded packets collected. Therefore, the gain induced by network coding may be dramatically decreased once a block cannot be decoded due to the arrival of PUs. In this paper, for the first time, we develop an efficient network coding strategy for SUs while considering the uncertain idle durations in CRNs. To handle the uncertainty of SUs' available transmission time, we first consider how to estimate the length of idle duration. For the case where the length of idle duration is stochastic, we employ confidential interval estimation (CIE) method to estimate the expected length of the idle duration. For the non-stochastic case, we utilize multi-armed bandit (MAB) to determine the idle durations sequentially. After obtaining the estimated length, we further adopt systematic network coding (SNC) in the data transmission of SUs. We find that SNC is more suitable for SUs' transmission than the general block-based network coding in the sense that it can reduce average perpacket delay without decreasing the throughput gain. However, the block size (also the proportion of uncoded packets to be sent) of SNC is hard to determine, due to the complicated correlation among the receptions at different receivers. To solve this problem, we propose an optimal block size selection algorithm for SNC (OSNC) to determine the transmission proportion of uncoded packets, under a given idle duration length. Due to its low computational complexity, OSNC can be used to make an online decision on the optimal block size with small delay. Simulation results show that, compared to traditional block-based network coding and plain retransmission schemes, our proposed scheme achieves highest performance for both stochastic and non-stochastic idle durations.