西南交通大学学报2026，Vol.61Issue(1)：156-166,11.DOI:10.3969/j.issn.0258-2724.20230670
基于外码分块编码的BATS码度优化
Degree Optimization of Batched Sparse Codes Using Outer Code Block Encoding
摘要
Abstract
To address the issues of repeated decoding and resource waste caused by random batch generation of the outer code in existing outer code block coding schemes for batched sparse(BATS)codes,the optimization of theoretical batch count and dynamic adaptability of BATS codes was systematically investigated based on the outer code block encoding scheme.First,under the condition of a known packet loss rate,a batch consumption analysis model for BATS codes was established,and an optimal degree value computation method was derived to tackle the challenges in existing schemes regarding theoretical batch count calculation and optimal degree value determination for minimizing batch count consumption.Second,for scenarios with unknown packet loss rates in the channel,a reinforcement learning-based dynamic degree optimization method for BATS codes was proposed,enabling real-time acquisition of degree values through an intelligent learning mechanism.Finally,simulation experiments were conducted to evaluate the theoretical model and the proposed dynamic optimization method.Simulation results have shown that the established transmission model based on outer code blocks and its batch count computation formula can be used to calculate batch consumption and determine the optimal degree distribution.Simulation results demonstrate that the proposed reinforcement learning-based optimization scheme achieves lower average batch count consumption than fixed-degree value schemes with unknown packet loss rates and maintains great performance in dynamic channel environments.关键词
分批稀疏码/分块码/传输次数/强化学习Key words
batched sparse(BATS)code/code block/transmission time/reinforcement learning分类
信息技术与安全科学引用本文复制引用
杨柳,阴慧颖,马征,刘恒,王士恒..基于外码分块编码的BATS码度优化[J].西南交通大学学报,2026,61(1):156-166,11.基金项目
国家自然科学基金项目(U2268201,62020106001) 轨道交通信息化国家重点实验室开放研究基金项目(SKLKZ22-02)支持. （U2268201,62020106001）
