Robust Model Reference Adaptive Controller Design

9413123考夫曼提出的各种直接模型参考自适应控制方法特别吸引人，因为它们不需要完整的状态反馈或自适应观测器。这类算法的其他重要性质包括(1)它们对非最小相位系统的适用性，(2)对象的阶数可能比参考模型的阶数大得多，以及(3)这种方法考虑具有多个输入和输出的对象。然而，值得注意的是，为了保证模型的渐近跟随甚至有界性，证明满足正实类型约束就足够了。虽然已经开发了各种增强方案来缓解这一限制，但设计所需的(通常是前馈)补偿以在期望的(或最大的)参数变化范围内满足正实条件并非易事。因此，建议考虑最近的稳健补偿技术，这些技术似乎与这种补偿的设计有关。其中包括为单输入单输出自适应系统开发的鲁棒前馈补偿器，它将扩展到多输入多输出系统，以及将被考虑用于单输入单输出系统的Kharitonvoc多项式分析。结合MRAC算法所得到的增广过程应该是设计和综合非常稳健的直接自适应控制器的容易实现的过程，这些控制器适用于许多具有显著不确定性的系统。***

9413123 Kaufman Various direct model reference adaptive control (MRAC) procedures developed by the proposed investigator is especially attractive because they do not require full state feedback or adaptive observers. Other important properties of this class of algorithms include (1) their applicability to non-minimum phase systems, (2) the order of the plant may be much larger than the order of the reference model, and (3) this approach considers plants with multiple inputs and outputs. However, it is important to note that in order to guarantee asymptotic model following or even boundedness, it is sufficient to demonstrate the satisfaction of a positive real type constraint. Although various augmentation schemes have been developed to alleviate this constraint, it is not straightforward to design the needed (usually feedforward compensation to satisfy the positive real condition over the desired (or a maximal) range of parameter variations. Thus, it is proposed to consider recent robust compensation techniques that appear relevant to the design of such compensation. These include a robust feedforward compensator developed for single input-single output adaptive systems which will be extended to multiple input-multiple output systems and the Kharitonvoc polynomial analysis, which will be considered for use in single input-single output systems. The resulting augmentation procedures from this proposed effort in conjunction with the MRAC algorithms should result in easily implementable procedures for designing and synthesizing very robust direct adaptive controllers that are applicable to many systems with significant uncertainty. ***