A novel adaptive control scheme is presented. It utilizes the inverse system concept along with the conventional model-following adaptive control scheme. Initially a basic control scheme for controlling linear time-invariant plants is considered. The scheme is then extended to the adaptive control of time-varying plants by incorporating periodical identification of the plant dynamics and updating its inverse. The algorithm used here is a computer-oriented adaptation of the Silverman's inversion algorithm. The software developed for calculating the inverse is appended. The main results can be summarized as follows:

(i) The plant can be made to follow the model to any desired degree of closeness in the presence of disturbance by suitably choosing a feedback gain parameter.

(ii) If certain conditions are satisfied, the model can supply the required information to the inverse without explicit differentiation.

(iii) When some higher-order derivatives of the plant output are not measurable, they can either be generated by using pseudo-differentiators or, in some cases, they may be ignored. In the former case, the frequency spectrum of disturbance-response attenuation is flat over all frequencies while in the latter, the spectrum is similar to that of the former case at low frequencies but the attenuation decreases above a certain frequency and reaches 0-dB level asymptotically.

(iv) When an inverse does not exist, the use of a pseudoinverse is suggested.

The scheme is illustrated by three examples, namely,

(i) design of a roll-attitude controller for a typical VTOL in hover.

(ii) design of a roll- pitch- and yaw-attitude controller for a typical VTOL in hover; here the design also includes decoupling.

(iii) control of the short-period approximation of longitudinal dynamics of a typical airplane during an accelerated flight.

In the first two examples, the plant has a nonchanging dynamics while in the last example, the dynamics changes with forward velocity. These examples were simulated using a digital system-simulation software package. The results are very encouraging. The first two examples were also simulated on a hybrid computer for the case of no disturbances. These results compare very well with those obtained digitally.

Among the main features of the proposed scheme are:

(i) Simplicity. No optimization is necessary. A substantial saving in computation time seems possible since the calculation of the inverse is quite straightforward. Moreover, the results are guaranteed since no search in an unknown parameter space is involved.

(ii) The scheme is more powerful than the commonly used (nondynamic) state-variable feedback control since it automatically incorporates, when necessary, the equivalent of any dynamic elements required in the state-feedback path.

(iii) Direct approach, i.e., the design of the controller starts with the generation of the desired output which is then functionally reproduced using an inverse.

(iv) The scheme provides a plant-independent method of designing a hybrid controller.

Further research possibilities include derivation of the inverse of non-linear and distributed-parameter systems and incorporation of any control constraints in the inverse-system formulation.