System Simulation
Estimating reliability in simulation experiments
Proceedings of the second conference on Applications of simulations
Hi-index | 0.00 |
A major problem in system simulation is the handling of random phenomena. The Monte Carlo method, which is usually applied for this purpose, is known to converge rather slowly. A different method is presented here which deals directly with probability distributions instead of random samples. Arithmetic operations on sampled values of random variables are replaced by transformations of their distributions. The main advantage of this method is that it is arbitrarily precise. Therefore, long runs for gathering statistics are not necessary. However, difficulties can arise from large memory requirements and program complexity. A comparison with the Monte Carlo method is given on the basis of two examples, the simulation of a signalized traffic network and of a supermarket.