In the context of fruit harvesting, the grasping force of the robotic end-effector is often unstable, leading to uneven force distribution on the fruit, which may either damage the fruit or cause it to slip. To address this issue, this study proposes an end-effector force control strategy based on Fuzzy Neural Network PID (FNN-PID). Starting from the voltage-to-torque transfer function, a unified transfer function model of the motor-transmission-end force system is established. On this basis, three types of controllers are designed: traditional PID, fuzzy PID, and FNN-PID, and their performances are compared under identical operating conditions. The results show that when the controlled object parameters remain constant and external disturbances are present, FNN-PID outperforms the other two controllers in terms of adjustment response time, overshoot, and steady-state error. Furthermore, reproducible Simulink implementation procedures and key parameter settings are provided using MATLAB, offering a theoretical foundation for the optimization design of robotic grippers.