The Design/Modelling/Performance Trade-Offs in High-Speed Robot Manipulation

Earlier work demonstrated that robust robot controllers based on simplified models greatly reduced controller complexity and enhanced physical insight, yet outperformed standard schemes based on `complete´ models at most practical levels of parametric uncertainty. In this paper, we detail some of the practical implications of robust robot controller design in terms of modelling/performance trade-offs. In particular we quantify parameter error sensitivity and the incremental effects of model simplification on controller performance. These issues are particularly critical in the case of direct-drive arms, which combine high-speed capabilities with full coupling and nonlinear effects.