How does the central nervous system select and coordinate different degrees of freedom to execute a given movement? The difficulty is to choose one specific motor command among an infinite number of possible ones. If some invari-ants of movement can be identified, there exists however considerable variability showing that motor control favors rather an envelope of possible movements than a strong stereotypy. However, the central nervous system is able to find extremely fast solutions to the problem of muscular and kinematic redundancy by producing stable and precise movements. To date, no computational model has made it possible to develop a movement generation algorithm with a performance comparable to that of humans in terms of speed, accuracy, robustness and adaptability. One of the reasons is certainly that correct criteria for the synthesis of movement as a function of the task have not yet been identified and used for motion generation. In this chapter we propose to study highly dynamic human movement taking into account its variability, making the choice to consider performance biomechanical variables (tasks) for generating motions and involving whole-body articulations.