Multi-Core Computing Cluster for Safety Fan Analysis of Guided Projectiles

摘要

During the design phase of Guidance, Navigation, and Control (GN&C) systems for guided projectiles, performance evaluation of the algorithms, and the resulting flight dynamics, is invaluable. In order to characterize the performance of a system's effectiveness, a Monte Carlo analysis is typically performed using empirically derived random variable distributions. Due to the complexity of the GN&C system, typical methods of linearization and closed form solution evaluation do not provide adequate results when determining the key parameters for measuring overall system performance. Since reliance on traditional methods is not adequate, Monte Carlo analysis must be performed with thousands, if not hundreds of thousands of iterations. Analysis of this complexity requires significant computational assets, therefore this process is used sparingly during the design phases. However, determining the true robustness of a control system cannot be resolved without performing this critical analysis. Through parallelization, this process can be run independently on many processors simultaneously, reducing the run time and allowing for quick evaluation of design modifications and algorithm updates. By using a small computing cluster and rapidly parallizing simulations, the Monte Carlo analysis can be integrated into the design cycle and mitigate errors prior to costly flight experiments. This rapid analysis provides a critical tool, not only for the evaluation of the system performance, but for safety fan analysis as well.