仿真系统主流开发平台随机数发生器的分析

摘要

随机数序列统计品质对蒙特卡洛仿真试验的精度和效率具有重要影响.给出了Visual C++、Borland C/C++、Delphi、MS Fortran、Matlab、SIMSCRIPT、Simula等仿真系统主流开发平台上标准随机数发生器算法的一般形式,认为:所生成序列统计品质具有周期较短、相关性较强、存在高维稀疏网格等缺陷；设计方法落后,辅助功能不足；应采用面向对象技术构建组合发生器,以满足当前多样本并行仿真试验的需求.%The statistics quality of random sequence affects greatly on the precision and efficiency of Monte Carlo ( MC) Simulation experiments. After giving a general descriptive form of algorithms for standard pseudorandom number generators ( PRNGs) of the mainstream simulation development platforms for Simulation applications, such as Visual C + + , Borland C/C + + , Delphi, MS Fortran, Matlab, SIMSCRIPT and Simula, the paper points out that their generated random sequences have bad statistical quality, e. G. , short period length, relatively strong correlation , and regular multidimensional lattice structure, and that their design methods lag behind the times and their supporting functions are insufficient. To achieve statistical quality of random sequence and to make up inadequate auxiliary utilities, the paper proposes that the object - oriented technology should be applied to implement the combined linear congruent generator to meet the requirements of multi - sample parallel MC Simulation experiments.