In engineering applications, many control devices have been developed to reduce the vibrations of structures. Active tuned mass damper system is one of these devices, which is a combination of a passive tuned mass damper system and an actuator to produce a control force. The main objective of this paper is to present a practical procedure for both deterministic and probabilistic design of the active tuned mass damper control system using multi-objective genetic algorithms to mitigate high-rise building responses. For this purpose, extensive numerical analyses have been performed, and optimal robust results of the active tuned mass damper design parameters with their effectiveness in reducing the example building responses have been presented. Uncertainties, which may exist in the system, have been taken into account using a robust design optimization procedure. The stiffness matrix and damping ratio of the building are considered as uncertain random variables; and using the well-known beta distribution, 50 pairs of these variables are generated. This resulted in 50 buildings with different stiffness matrices and damping ratios. These simulated buildings are used to evaluate robust optimal values of the active tuned mass damper design parameters. Four non-commensurable objective functions, namely maximum displacement, maximum velocity, maximum acceleration of each floor of the building, and active control force produced by the actuator are considered, and a fast and elitist non-dominated sorting genetic algorithm approach is used to find a set of pareto-optimal solutions.