Abstarct: This thesis, a balance recovery for the humanoid robot of the Nao-21 is improve. the objective function baxsed on the predictive model control approch and evaluation. Considering the position of capture point and position of zero moment point and the angular momentum rate of robot and position of the robot as the system mode vectors and the rate of speed changes of the trunk and zero moment rate changes as a control input is an optimization issue is designed. A combined ankle and hip strategy has been used for the recovery of the balance of other strategies and have been investigated. Considering the inverse pendulum dynamic model, along with with with with the rotary cycle and zero moment point modulation and the rotational torque of the trunk is a limited and definite cost function that forms the capture point. By applying external disturbance, the capture point has been supported outside the support polygon, which with the help of the predictive controller of the cost function in the most optimum situation, and by applying changes in the rotating torque of the trunk capture point to the support polygon was guided and the recovery of balance of the robot is possible. All of these simulations were performed in MATLAB software and were used to evaluate the software emulator. Finally, the empirical implementation of this algorithm has been done on the real robot. Due to hardware weakness of Nao robot in the pressure sensor part of a three-dimensional procedure, which includes the barometer element, is designed and manufactured for evaluation in MATLAB software and features of the sensor outputs and the designed elements are compared to the results of efficient elements And with good accuracy, it has been concluded that in this study, it has resulted in improved performance of robot balance recovery.