This paper presents the oxygen stoichiometry control problem of proton exchange Membrane (PEM) fuel cells and introduces a solution through an optimal control methodology. Based on the study of a non-linear dynamical model of a laboratory PEM fuel cell system and its associated components (air compressor, humidifiers, line heaters, valves, etc.), a control strategy for the oxygen stoichiometry regulation in the cathode line is designed and tested. From a linearized model of the system, an LQR/LQG controller is designed to give a solution to the stated control problem. Experimental results show the effectiveness of the proposed controllers design.