The clogged-loop and human-robot method requires an effective way of robotic controller which considers the model that the robot and human, as well as to adaptation of human to robot.This work aims to develop a mid-level oriented controller that provides assist-as-needed (AAN) method in a hierarchical controlling set up by two methods as model- oriented and the fuzzy based rule. AAN is to provide the required additional torque since the robot's dynamics and its heavy load will be compared to that of the human limb with free movement. Human-robot adaptation is be simulated by a nonlinear method predictive based controller (NMPC) since the human ways of central nervous system has three conditions of the initial (starting of wearing robot, without old experience), short-term, and long-term experiences. A method of experiment that results in two models (fuzzy and model-based) that outperforms the earlier proportionals model by using AAN which considers the distinctive human and the robot models. And also the CNS model is difficult at the initial level experience and also activates both the antagonist and the agonist muscles to minimize a movement based oscillations. Due to long-term based experience, the simulation results shows no oscillation and when the CNS NMPC learns robot model and changes its weights so as to simulate the realistic of human behavior. Its is found that the preferred strength of robot remains increased gradually in order to ignore the unwanted human-robot interaction (e.g., human spasticity, robot vibration, ). This proposed mid-level based controllers can be used to wearable assistive devices, exoskeletons, and rehabilitation robots applications

,