Adaptive and Robust Nonlinear Control

小行星9528370 这是一项关于继续我们的NSF资助#ECS-9203491的提案，该资助将于1995年8月31日到期。 提案的第一部分总结了我们目前资助的成果，这些成果已在一本550页的专著中提出，并在20多篇期刊论文和更多的会议论文中提出。 我们的主要成就之一是新的递归自适应和鲁棒设计程序（“backstepping”）的非线性系统的参数和功能的不确定性。 对于我们的自适应设计，我们推导出的性能界限，在线性系统的情况下，表明它们优于传统的自适应计划。 虽然我们目前的研究结果已经非常好地接受了研究界，也激发了应用的兴趣，我们的批判性审查其局限性的动机，这一建议的主要议题：第一，继续发展从可行到可实施的设计，第二，开发新的设计程序更大的系统类。 我们还建议启动一个深入的研究结构特定的非线性反馈回路的鲁棒性，特别是那些诊断我们的自适应程序。 为了使我们的递归程序更有吸引力的应用程序，我们建议开发精益反推设计，避免浪费的取消，减少控制工作。 为此，我们将进一步发展我们的逆优化设计。 为了扩展类的系统，我们计划开发新的转发程序作为上三角模拟的backstepping。 我们将扩展的不确定性的应用动机的“形状”的不确定性和开发“形状鲁棒”控制器的特征。 我们的鲁棒性研究将量化性能鲁棒性的权衡，并开发重新设计，以提高鲁棒性，同时最大限度地减少性能损失。 NSF支持的研究为我们与联合技术公司、福特汽车公司和罗克韦尔国际公司的工业合作提供了理论基础和动力。 ***

9528370 Kokotovic This is a proposal for continuation of our NSF Grant #ECS-9203491 which expires on August 31, 1995. Part one of the proposal summarizes the results under our current grant which have been presented in a 550-page monograph, and in over 20 refereed journal papers and more numerous conference papers. Among our major accomplishments are new recursive adaptive and robust design procedures ("backstepping") for nonlinear systems with parametric and functional uncertainties. For our adaptive designs we have derived performance bounds which, in the case of linear systems, show that they outperform traditional adaptive schemes. Although our current results have been extremely well received by the research community and have also stimulated interest in applications, our critical examination of their limitations has motivated the main topics of this proposal: first, to continue to the development from feasible to implementable designs, and, second, to develop new design procedures for larger classes of systems. We also propose to initiate an in-depth study of structure-specific robustness properties of nonlinear feedback loops, especially those diagnosed by our adaptive procedures. To make our recursive procedures more attractive for applications, we propose to develop lean backstepping designs which avoid wasteful cancellations and reduce control effort. For this purpose we will further develop our inverse-optimal designs. To extend the classes of systems, we plan to develop new forwarding procedures as upper-triangular analogs of backstepping. We will extend characterizations of uncertainties by application-motivated "shape" uncertainties and develop "shape-robust" controllers. Our robustness studies will quantify performance-robustness trade-offs and develop redesigns to enhance robustness while minimizing the loss of performance. Our NSF-supported research provides a theoretical base for, and is motivated by, our industrial cooperation with United Technologies, Ford Motor Company and Rockwell International. ***