For the safety and efficacy of in vivo therapy using electromagnetic actuated microrobots, position tracking performance is crucially important. However, it is difficult to manipulate the microrobots accurately and rapidly due to the nonlinear and complex dynamics of the microrobots and highly limited magnetic force of the actuator. In this paper, we propose a robust control law for the microrobots. It consists of a time-delay estimation to compensate for the unknown/unmodeled dynamics, an antiwindup scheme, and a forgetting factor to improve the performance. In addition, we added a switching action to the control law in order to guarantee the stability of the controller. The improved performance and stability of the proposed control law were verified through an experiment that contains 3-DOF motion.