An adaptive control scheme may provide the best approach to the problem of accommodating automatic flight control systems to the variations of dynamic characteristics encountered over the flight envelope of the aircraft. The C*-Criterion for aircraft time response provides a basis for the design of such a system. To provide a margin of safety, all control systems have redundant channels for emergency use. An adaptive control system can be designed to be self-organizing, and in such a configuration can provide its own failure monitoring, resulting in a simpler and more efficient system. This study shows that such a system is feasible, that the response of the system is within the limits set forth in the C*-Criterion, and that the self-organizing characteristics provide reliable operation over the whole range of flight operations. (Author).

A study of the feasibility of a self-learning adaptive system for the flight control of high performance aircraft has been performed. A flight control system was developed for the investigation of the stability augmentation of the longitudinal axis of the F101B aircraft using self-learning adaptive control. The learning adaptive controller developed employs a three-loop concept. The innermost loop comprises a linear feedback control system in which a set of control gains is adjusted by a second (adaptive) loop employing a parameter identifier and a trainable function generator (automation). The automation provides the correct values of feedback gain in response to patterns derived from the identified aircraft parameters. The third loop (the learning loop) measures control system performance, and continually retrains the automation to improve the performance. Experiments were performed with a digital simulation of the aircraft and the learning adaptive control system. Results of the experiments indicate that learning adaptive control is feasible. However, a number of significant technical problems must be overcome prior to the use of such a control system in tactical aircraft. Studies of sensitivity must be performed to assess the effect of small perturbations in the identification parameters on system performance. Extended studies of property extraction from the identified parameters is required, and means for further simplifying the control structure is of importance in a real system. (Author).