With the expanding applications of mobile robots, precise positioning and orientation have become key techniques for achieving autonomous navigation and control. However, due to the complexity of the environment and sensor errors and other factors, there are certain biases in the estimation of the posture of mobile robots. For this reason, a digital twin system is built for mobile robots based on the ROS, and a virtual-real positioning error compensation method is proposed under this system. An Adaptive Genetic Algorithm-based Predictive Optimization Controller (AGA-POC) is designed to drive the controlled models by issuing control commands to compensate for the positioning error existed in the virtual-real mapping process. The experimental results show that compared with Model Predictive Control (MPC) compensation, the positioning error and trajectory deviation of the mobile robots are reduced by 60.43% and 35.66% respectively, the trajectory similarity and controller response rate are improved by 69.28% and 71.14% respectively, which verifies the feasibility and validity of the compensation strategy in terms of positioning accuracy, trajectory accuracy and controller response rate.