Self-adaptive Parameter Research Articles

Evolutionary algorithms

Genetic Algorithms and Evolution Strategies, the main representatives of a class of algorithms based on the model of natural evolution, are discussed w.r.t. their basic working mechanisms, differences, and application possibilities. The mechanism of self-adaptation of strategy parameters within Evolution Strategies is emphasized and turns out to be the major difference to Genetic Algorithms, since it allows for an on-line adaptation of strategy parameters without exogenous control. Furthermore, the importance of selection as a mechanism to control the character of the search to be either more volume oriented or more path oriented is pointed out.

Phase and amplitude sinusoidal dither adaptive control system

In many control system applications an adaptive system is required to accommodate changes in plant dynamics. A self-adaptive system described herein adjusts two control system parameters to maintain the closed-loop response essentially constant in spite of large changes in the plant gain and pole-zero locations. The adaptive system uses a high-frequency dither frequency to interrogate the state of the control system. Two self-adaptive performance criteria (SAPC) adjust the two adaptive parameters. One SAPC is based upon the amplitude response to the dither frequency and the other is based upon the phase response. The adaptive loop associated with the amplitude SAPC adjusts the control system loop gain; the adaptive loop associated with the phase SAPC adjusts a variable compensation network. A method is developed for analyzing the stability and performance of the self-adaptive parameter adjustment loops. The stability analysis includes the interaction between these adaptive loops. This two-parameter adaptive system is applied to a specific example to illustrate the techniques developed in the paper.