With the development of aerospace technology, more and more space manipulators are being applied to orbit services. In order to ensure the stable operation of the space manipulator in space, it is necessary to perform a large number of simulation experiments on the ground microgravity environment to verify its various performance indicators before launching. In order to realize the ground microgravity environment of the space manipulator, and at the same time ensure the high-precision microgravity environment simulation effect, a three-dimensional active suspension microgravity simulation system is designed in this paper. The system is provided with 3 lifting point mechanisms, and the mechanism for microgravity simulation of each hanging point mainly includes a constant tension hanging unit and a 2-dimensional horizontal linear motion unit. The constant tension suspension unit adopts the active control mode to realize the motion following and vertical gravity compensation of the space manipulator through the feedback of the force sensor. The control accuracy of this unit directly affects the accuracy of gravity compensation. The 2-dimensional horizontal linear motion unit uses the feedback of the angle sensor to actively follow the movement of the space manipulator in the horizontal direction through active control. The motion following accuracy of this unit determines whether horizontal component force will be introduced and affect the accuracy of gravity compensation. This article mainly studies how to improve the accuracy of gravity compensation of suspended microgravity simulation system. First, a mathematical model is established for the constant tension hanging unit, and a constant force / position hybrid control algorithm based on fuzzy PID parameter tuning is proposed using the constant tension control idea. The algorithm adopts a structure in which the position control loop and the force control loop are controlled in parallel. In the force control loop, the PID controller parameters are adjusted online through fuzzy control to achieve high-precision gravity compensation. A composite control strategy for the combined motion of the hoisting motor and electric cylinder of the unit is proposed to realize the extension of the displacement range of the electric cylinder by the hoisting motor. Aiming at the 2D horizontal linear motion unit, an incremental PID control algorithm based on RBF neural network system identification based on 3-8-1 network structure is designed. The adaptive adjustment of the PID controller parameters is realized through the RBF neural network to improve the control accuracy of the following motion in the horizontal direction, so as to avoid the influence of the introduction of horizontal component force on the accuracy of gravity compensation. In addition, this paper also gives the hardware composition and selection of the control system and the design flow of the software functions. A simulation model is built in the MATLAB / Simulink environment. Simulation results show that the hybrid force / position control algorithm based on fuzzy PID parameter tuning can control the accuracy of gravity compensation within 0.3% F.S (full range). The PID control algorithm based on RBF neural network system identification can control the deflection angle of the sling within 0.4 °. And the entire gravity balance system has strong anti-interference ability and good dynamic response performance. Coupled with the data obtained from the actual field experiments, it can be seen that the control algorithm designed in this paper meets the project index requirements of the suspended microgravity simulation system.