Abstract
We propose a method to analyze underwater robots basing on a learning control and a nonlinear time scale translation. In the water, the additional forces and moments called added mass, hydrodynamic damping and buoyancy act on the arm and it is difficult to estimate these forces and moments. At first, the input torque pattern of the learning control is modified by the tracking error without estimating added mass, hydrodynamic damping and buoyancy. Next, the obtained torque pattern is investigated through a nonlinear time scale translation.
