This paper presents the design and optimization of a new alternating electromagnetic-field-generation system, which is dedicated to actuating untethered magnetic microrobots under an inverted microscope. Its uniqueness is that the system parameters are optimally designed by considering both electric and geometry constraints for the target-driving application. The dominant parameters of the system are first determined by establishing analytical models. According to the requirements of targeted application, the optimization problem with certain constraints is formulated, which is solved via the multiobjective genetic algorithm method. A prototype system with the optimal parameters is developed for experimental testing. Experimental studies are carried out to characterize actual performance of the developed actuation system. For demonstration, a magnetic microball has been actuated for navigation by surface rolling in a petri dish filled with pure water. Results indicate that the reported electromagnetic-field-generation system meets the actuation requirements for potential applications.