Intelligent Control Systems Design

控制系统需要智能才能在具有高度不确定性的动态环境中有效运行。在这种情况下，成功的自主操作可能源于精心设计的各种信息处理能力的互连。这种能力必然意味着设计师对于哪些信息是重要的以及它如何产生控制行为的看法的某种计算表示。因此，智能控制器产生了在扩展状态空间上定义的动态系统，其底层算法相当于在这些空间中进行调谐和切换的规则。我们建议探索一个统一的数学框架，以稳定的方式组合在线设置中的各种信息处理能力。我们打算详细探讨两个具体的具有挑战性的应用-飞机和机器人控制问题-为我们的理论研究提供刺激。最后，我们计划设计一个统一的架构，通过系统地使用人工神经网络来实现不同的信息功能。

Control systems require intelligence to operate efficiently in dynamical environment possessing a high degree of uncertainty. Successful autonomous operation in such situations can result from a carefully designed interconnection of diverse information processing capabilities. Such capabilities necessarily imply some computational representation of the designer's view of what information is important and how it can produce the desire control action. Thus intelligent controllers give rise to dynamical systems defined on extended state spaces and their underlying algorithms amount to rules for tuning and switching in those space. We propose to explore a unified mathematical framework for combining, in a stable manner various information processing capabilities in on-line settings. We intend to explore in detail two specific challenging application - an aircraft and a robot control problem - to provide the stimulus for our theoretical investigations. Finally, we plan to devise a uniform architecture for implementing the diverse information capabilities by the systematic use of artificial neural networks.