Model-adaptive hybrid dynamic control for robotic assembly tasks

A new task-level adaptive controller is presented for the hybrid dynamic co ntrol of robotic assembly tasks. Using a hybrid dynamic model of the assemb ly task, velocity constraints are derived from which satisfactory velocity commands are obtained. Due to modeling errors and parametric uncertainties, the velocity commands may be erroneous anti may result in suboptimal perfo rmance. Task-level adaptive control schemes, based on the occurrence of dis crete events, are used to change the model parameters from which the veloci ty commands are determined. Two adaptive schemes are presented: the first i s based on intuitive reasoning about the vector spaces involved whereas the second uses a search region that is reduced with each iteration. For the f irst adaptation law, asymptotic convergence to the correct model parameters is proven except for one case. This weakness motivated the development of the second adaptation law, for which asymptotic convergence is proven in al l cases. Automated control of a peg-in-hole assembly task is given as an ex ample, and,simulations and experiments for this task are presented. These r esults demonstrate the success of the method and also indicate properties f or rapid convergence.