A study on a pan-tilt type target tracking system actuated by permanent magnet DC motors and assembled in a moving body is presented in this work. To achieve such objective, an experimental test bed is constructed and a simulation program is implemented. The mechanical model is derived and simulated in time domain. This approach allows using accurate non-linear equations to represent system behavior, otherwise infeasible in frequency domain. Although the system is modeled with rigid bodies, flexibility and structural damping due to the gearboxes are considered. Sensor errors, backlash in the gearboxes, dry and viscous friction, saturation limits for armature current and tension of the motors are also considered. A method to include the time delays for the control signal updates, as well as time delays due to sensor dynamic response, during the numerical integration of the equations of motion, is presented. Controllers that require no mathematical model of the plant are employed in the experimental test bed and in the simulation program. Three different control architectures are proposed, called in this work type 1, type 2 and type 3. Their complexity increases depending on the number of available sensors. The type 1 is applied to systems with only one sensor that provides the targets angular azimuth and elevation errors. If, besides this sensor, sensors to measure the relative angular positions between the mechanism links are available type 2 architecture is used. In addition, if sensors to measure inertial angular speeds are also available, type 3 architecture can be used. Finally, experimental and numerical results, comparing system performance with each control architecture are presented.