A horizontal stability optimization control strategy of port overload AGV under low attachment coefficient road surface was proposed based on three-layer control structure. The MPC algorithm was used in the upper controller, which can continuously sample and predict the AGV dynamic parameters and alleviate the time lag problem of the control system caused by the series connection with the middle and lower-level controllers. The genetic algorithm was used in middle-level controller to optimize the parameters of the fuzzy controller for application to the stable control of the port AGV body. The SQP algorithm was applied in the lower-level controller to convert the driving torque optimal distribution problem into a quadratic programming sub-problem  and find an online solution. The Matlab / Simulink and Trucksim joint simulation platform based on port AGV was constructed to verifies the validity and robustness of the proposed strategy.