Quasi-Universal BATS Code
Xiaoli Xu, Yong Liang Guan, Yong Zeng, Chee-Cheon Chui · IEEE Transactions on Vehicular Technology · 2016
Batched sparse (BATS) code is a class of temporal network code that achieves near-optimal tradeoff between network throughput and coding length for multi-hop erasure networks. However, the performance of the traditional BATS code degrades dramatically when the designed degree distribution does not match the actual channel condition. In this paper, we first prove that a universal degree distribution that asymptotically achieves the optimal rate for all channel rank distributions does not exist for BATS code with batch size greater than one. We then propose a quasi-universal BATS (QU-BATS) code that achieves near-optimalperformance for a range of channel conditions. This makes it suitable for use in scenarios where the end-to-end channel rank distribution is not fixed or not exactly known, e.g., in multicast or wireless transmission. In the proposed QU-BATS coding scheme, multiple degree distributions are designed, and the coded packets are generated according to different degree distributions at different transmission stages. Simulation results show that the proposed QU-BATS code strictly outperforms the fountain code and the traditional BATS codes for multihop data streaming over uncertain or time-varying network links, with lower decoding complexity.