Some properties of fuzzy relational matrix (Q2735366)

The present paper is devoted to the study of fuzzy logic controllers. Let \(a,b\in [0,1]\). The authors propose the following new continuous \(T\)-norm and \(S\)-norm which vary through different operators when the correlation coefficient \(h\in [-1,1]\) is changing from \(-1\) to \(1\):NEWLINENEWLINENEWLINENEWLINE\[NEWLINET_h(a,b)= \begin{cases} (1- h)T_P+ hT_Z,\quad &0\leq h\leq 1,\\ (1+ h)T_P- hT_L,\quad &-1\leq h\leq 0;\end{cases}NEWLINE\]NEWLINE NEWLINE\[NEWLINES_h(a,b)= \begin{cases} (1- h)S_P+ hS_Z,\quad & 0\leq h\leq 1,\\ (1+ h)S_P- hS_L,\quad &-1\leq h\leq 0;\end{cases}NEWLINE\]NEWLINE where \(T_Z= \min(a,b)\), \(S_Z= \max(a,b)\) are Zadeh operators, \(T_P= ab\), \(S_P= a+ b-ab\), are the probability operators and \(T_L= \max(0,a+ b-1)\), \(S_L= \min(1, a+b)\) are Łukasiewicz operators. To replace conventional MAX/MIN operators, the following simplified multi-input AND/OR operators are suggested in the study of fuzzy logic controllers: For \(a_i\in [0,1]\), \(i= 1,2,\dots, k\), NEWLINE\[NEWLINET_h(a_1,\dots, a_k)= \begin{cases} (1- h)a_1\cdots a_k+ h\min(a_1,\dots, a_k),\quad &0\leq h\leq 1,\\ (1+ h)a_1\cdots a_k- h\max(0, a_1+\cdots+ a_k- k+1),\quad &-1\leq h\leq 0;\end{cases}NEWLINE\]NEWLINE NEWLINE\[NEWLINE\begin{multlined} S_h(a_1,\dots, a_k)= \\ = \begin{cases} (1- h)[1- (1- a_1)(1- a_2)\cdots(1- a_k)]+ h\max(a_1,\dots, a_k),\;0\leq h\leq 1,\\ (1- h)[1- (1- a_1)(1- a_2)\cdots(1- a_k)]- h\min(1, a_1+\cdots+ a_k),\;-1\leq h\leq 0.\end{cases}\end{multlined}NEWLINE\]NEWLINE The correlation coefficients can be decided by using the genetic algorithm. A simulation example is given to show that the method obtained is better than the conventional Mamdani method.