A Massively Parallel Bayesian Approach to Planetary Protection Trajectory Analysis and Design
暂无分享,去创建一个
The NASA Planetary Protection Office has levied a requirement that the upper stage of future planetary launches have a less than 10-4 chance of impacting Mars within 50 years after launch. A brute-force approach requires a decade of computer time to demonstrate compliance. By using a Bayesian approach and taking advantage of the demonstrated reliability of the upper stage, the required number of fifty-year propagations can be massively reduced. By spreading the remaining embarrassingly parallel Monte Carlo simulations across multiple computers, compliance can be demonstrated in a reasonable time frame. The method used is described here.
[1] Ian Foster,et al. Designing and building parallel programs , 1994 .
[2] B. Tapley,et al. Statistical Orbit Determination , 2004 .
[3] L. Brown,et al. Interval Estimation for a Binomial Proportion , 2001 .