借助概率母函数的分析方法，对输入缓存方式交换网络的性能进行了理论分析，扩展了前人的分析结果，推出了吞吐率和延时表达式的闭和形式；并对这种网络的输出竞争，提出一种简单的基于时延相关策略的神经网络仲裁算法，解决信元调度问题．仿真结果表明，该仲裁方法改善了网络性能． By using the analysis method of probabilistic parent function, the performance of the input buffer-based switched network is analyzed theoretically, which extends the previous analysis results and introduces the closed form of the throughput and delay expressions. And the output of this network Competition, propose a simple neural network based on delay-related strategy arbitration algorithm to solve the cell scheduling problem. Simulation results show that the arbitration method improves network performance.