Two approaches are proposed to model and control a human-like motion of robot arms. The first, which is based on the concept of distributed positioning (DP), is suggested as a good model of arm motion in the phase where fatigue does not appear. The prescribed motion of the end-effector is distributed to a redundant number of arm joints in accordance to their acceleration capabilities. For the phase where fatigue appears, the concept of virtual fatigue is proposed. This artificial variable, which is based on robot dynamics, emulates the progress of biological fatigue. The human handwriting task is chosen for the simulation. The DP concept is tested first by modeling nonfatigued motion. The justification of the usual inclination of letters is presented, and the relation between the inclination, legibility, and a secondary objective (finger involvement, energy consumption, motor thermal load) is discussed. It is found that, for some prescribed level of legibility, the individual optima of all the secondary cost functions are quite near to each other. Writing in the presence of fatigue is also analyzed, applying the method of the so-called "virtual fatigue".