Abstract: Abstract: In order to minimize the harm due to robot end-effector grasp fruits and vegetables, a generalized proportional integral (GPI) output feedback control scheme is proposed in this paper. First, the study object is the end-effector with two fingers that mounted on the top of a rigid manipulator, force sensors are mounted at the root of the finger, the output force signal was obtained by a force sensor, treating the model of the end-effector as a particle model and gravity does not affect the dynamics of end-effector, so the effect of gravity can be disregarded in the modeling. The modeling of the motor and the end-effector respectively and combining them together, the integrated model of end-effector is obtained, the mathematical relationship between the motor input voltage, and the load torque is also obtained. Then, the designed integral reconstructor, taking ramp error into consideration, the system gains robust characteristics to the ramp error that consists of force tracking error and constant disturbance, the integral reconstructor does not change the characteristic of the closed-loop. The computation of time derivatives can cause many common problems, such as a noisy signal, output delays, etc. In order to avoid these problems, making full use of integral reconstructor to design GPI force feedback controller, the force deviation converted to the input voltage control of the motor, this way needn't to derivate torque tracking error, so avoiding system time-delay and noise problem results from derivation. Making use of simulation software of MATLAB, taking sine signal and ramp signal as reference forces respectively, the force trajectory tracking error and the force output are analysed, the simulation result shows that the unit of force tracking error is 10-3, the real output force trajectory is close to the reference force trajectory, only having a little disturbance in the beginning, the output force trajectory reach a steady tracking status within 1 s, then the real output force trajectory smoothly tracks the reference force trajectory. Simulation results show that the GPI torque control has good torque tracking capability, which can decrease the damage of fruits and vegetables caused by end-effector grasping, and it is suitable for compliance grasp control of fruit-vegetable. Generalized proportional integral controller was applied to a fruit-picking robot to grasp apples and pears, after several the experiment of grasping fruits, the experimental results indicated that the method has a good performance and the fruit-picking robot can grasp and release fruit steadily and harmlessly. The ratio of successful grasping is 90%. Comparing with the situation that adopted traditional proportional integral control, the damage of fruits and vegetables caused by the end-effector grasping is reduced, fulfilling the demand of fruit picking. The research can provide a reference for the nondestructive picking of fruit-vegetable picking robot.