This article introduces an adaptive controller for a class of nonlinear discrete-time systems, based on self adjustable networks called Multi-Input Fuzzy Rules Emulated Networks (MIFRENs), and its reinforcement learning algorithm. Because of the universal function approximation of MIFREN, the first MIFREN called MIFREN(c) is used to estimate a long-term cost function, which demonstrates as a performance index for the tuning procedure. Another network or MIFREN(a) is designed as a direct controller via the human knowledge through defined If-Then rules. The selection procedure for any system parameters, such as learning rates and some constant parameters, is represented by the proof of proposed theorems. The system's performance is demonstrated by computer simulations via selected nonlinear discrete-time systems, and comparison results with other controllers to validate theoretical development.