EAGER: Control versus Prediction in Systems Engineering

系统工程是协调大型系统开发的所有部分的技术活动。 迄今为止，大型复杂工程系统（如飞机、航天器或运输系统）开发的系统工程管理的主要部分是基于预测开发项目的结果：需要多少时间？开发成本是多少？或者，最终产品的制造成本是多少？ 在目前的实践中，回答这些问题几乎总是导致低估。 尽管为改善此类预测进行了数十年的研究，但尚未取得任何进展。 该项目将试图找到一种更成功的方法来管理这些系统，不依赖于预测，而是专注于控制，控制成本，重量，进度和性能。 这是一个短期的、高风险的探索性研究项目，目的是建立一个以控制为基础的管理战略的可行性。 如果这一战略是可行的，那么就有必要进一步研究开发新的方法。研究计划是解构系统工程的本质，并根据现代控制理论中可用的方法和思想重建一种方法，然后确定如何测量这种新方法的性能，并利用其他领域有前途的研究方法。 然而，通过案例研究实际衡量该方法的绩效将留待今后的工作。 该项目将采用基本的最优调节器作为一种策略，并将探索一种类似动态编程的方法作为第二种方法。 如果成功的话，这些结果将为系统工程提供一种新的、更有效的方法。

Systems engineering is the technical activity that coordinates all the parts of a large system development. To date, a major part of systems engineering management of the development of large complex engineered systems such as aircraft, spacecraft or transportation systems has been based on predicting the outcome of the development project: How much time it will take? How much will the development cost? Or, how much will the eventual product cost to manufacture? In current practice, answering these questions almost always results in underestimates. Despite decades of research on improving such predictions, no progress has been made. This project will attempt to find a more successful approach to managing these systems that does not rely on prediction, but instead focuses on control, controlling cost, weight, schedule and performance. This is a short, high-risk exploratory research project to establish the feasibility of a controls-based management strategy. If the strategy is feasible, further research will be necessary to develop the new method.The research plan is to deconstruct systems engineering to its essence and reconstruct an approach based on methods and ideas available in modern control theory, and then determine how to measure the performance of this new approach, exploiting promising research methodologies from other fields. However, actual measurement of the performance of the method through case studies will be left for future work. The project will employ a basic optimal regulator as one strategy, and will explore a dynamic-programming-like approach as a second approach. The results will, if successful, provide insights into a new and more effective approach to systems engineering.