Model-based constrained motion. b. Introducing on-line identification and observation of the motion constraints

For pt.a see ibid., p.976-81 (1991). The authors present the fundamentals of a framework for model-based online identification of 'complex' motion constraints during force controlled 'constrained', or 'compliant', motion tasks. The identification approach is a generalization of the 'tracking' in Mason's 'compliance' or 'task' frame. Its range of possible applications, however, is much wider. Moreover, for the observation of the motion constraint state (i.e., the interpretation of the measured reaction wrench) nonstatic Coulomb friction is explicitly taken into account. The discussion uses the concepts 'feature' and 'virtual manipulator', not only to describe the motion constraint models, but also to define the uncertainty parameters and the identification and observation problems. Indeed, in this context identification is equivalent to finding the joint motions of the virtual 'uncertainty' manipulators in all features, while observation is equivalent to finding the joint torques of the 'reciprocal' virtual manipulator. Nontrivial, or 'complex', constrained motion tasks require an online 'task executor' to plan and schedule the sensing and identification actions. The text describes a very simple application of the presented 'feature-based identification and observation formalism': incompatible seam tracking.<<ETX>>