Dealing with unexpected situations during the execution of robotmotions

摘要

The problem of a polygonal object moving in an uncertain polygonalenvironment, i.e. the positions and orientations of the objects relativeto each other are not exactly known, is addressed. Furthermore, themoving object does not move exactly as it is commanded to move. Due tothese uncertainties, even a motion generated by a correct motion plannermay fail, i.e. for instance, an unexpected contact may occur. Asituation analyzer that enables the moving object to acquire informationon a contact situation is presented. This analyzer uses the nominalworld model including information on the magnitude of the uncertaintiesin order to get hypotheses on the geometric entities involved in thecontact. The discrimination among the hypotheses is based on testmotions, i.e. motions whose feasibility depends on the validity of acertain subset of the hypotheses