Self-tuning control has been recognised as an effective approach for mechanical manipulator control design due to its ability to cope with the presence of nonlinearities and uncertainties in robot dynamic models. The vast majority of existing self-tuning controllers are based on a linear plant description, the fact that most industrial processes are nonlinear is taken into account by regarding the plant as a sequence of pseudolinear descriptions. Therefore, as the plant operating point changes, the nonlinear plant dynamics are reflected as time varying parameters in the linear plant description. The applicability of this approach has been demonstrated by Koivo and Guo [1] where manipulator joint angular position is the controlled variable. This work is extended in Koivo et al [2] to the case in which the manipulator is controlled directly in the cartesian coordinate system. It is found that convergence of the parameter estimates may not be achieved during the finite time over which the· motion takes place. Therefore this approach is particularly suited to repetitive tasks where the last estimates from the previous run can be used as the initial estimates. Lelic and Wellstead [3] have successfully applied generalised pole placement to the control of a 5 axis electrically actuated robot-manipulator.