Computationally efficient dynamic modelling of robot arms (Q1905674)

Two algorithms are proposed yielding simplified, computationally efficient dynamic models of robot arms suitable for on-line identification and control. The simplification process for an \(n\)-link manipulator, assuming rigid body motion and excluding backlash, is performed off-line. Only position-dependent elements of the robot-arm models are approximated. Both trajectory-specific and general motion simplification options are presented. The first algorithm determines the set of fundamental parameters of an arm and eliminates other parameters that have no significant influence on the dynamic model. The second algorithm approximates each position-dependent element of the exact model by linear combination of simple functions of joint variables and is based on the Gram-Schmidt orthogonalization process.