|
To implement the response function of Fig 2, we use the input membership functions shown in (a) and the output membership singletons in (b). As with the P and D component, we normalize both domains to±1.0. Input functions are narrow to achieve good resolution across the domain-the increased number of rules is not important in a one-dimension rulebase. We then space the output singletons to approximate the response function. The rule "matrix" in (c) shows the input values above the cells and the corresponding output within each cell.
|
