Objective To study the control problem in dynamic foot biomechanical simulator and propose a complete multi-axis control algorithm which could be more competitive than that of current gait simulators in aspects as simulations in degree-of-freedom (DOF), velocity, precision, weight-bearing and trial efficiency. Methods A novel custom-made foot and ankle biomechanical simulator was developed to simulate both motion and force characteristics in a stance phase with 5 DOF. A model of the simulator was built in Matlab based on gait analysis and reasonable simplification. Iteration learning control (ILC) was proposed to control multi-axis forces and was verified in Simulink. Finally, the control strategy was validated in the simulation platform with a prosthetic foot. Results The novel simulator could complete the motion and force loading process within 5 seconds in one stance after 4-5 iterations. All 3D ground reaction forces (Fz, Fy and Fx) had high verified repeatability. The tracking curves of Fz and Fy with 50% of real body weight could converge to the target ones with root mean square (RMS) error of 20 N and 8 N using ILC, respectively, which was smaller than 10% of simulated loads. Conclusions The proposed control strategy greatly improved intelligence of the simulator and provided a good foundation to further improve the simulation speed and accuracy. The development of the simulator is of great significance to the cadaveric experiments on foot and ankle biomechanics.