A program of research in robotics that seeks to encode abstract tasks in a form that simultaneously affords a control scheme for the torque-actuated dynamical systems, as well as a proof that the resulting closed-loop behavior will correctly achieve the desired goals, is reviewed. Two different behaviors that require dexterity and might plausibly connote 'intelligence' -navigating in a cluttered environment and juggling a number of otherwise freely falling objects -are examined with regard to similarities in problem representation, method of solution, and causes of success. The central theme concerns the virtue of global stability mechanisms. At the planning level they lend autonomy, that is, freedom from dependence upon some 'higher' intelligence. They encourage the design of canonical procedures for model problems, which may then be instantiated in particular settings by a change of coordinates. The procedures developed result in provably autonomous behavior. Simulation results and physical experimental studies suggest the practicability of these methods. This material is posted here with permission of the IEEE. Such permission of the IEEE does not in any way imply IEEE endorsement of any of the University of Pennsylvania's products or services. Internal or personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution must be obtained from the IEEE by writing to pubs-permissions@ieee.org. By choosing to view this document, you agree to all provisions of the copyright laws protecting it.
AbstractAn autonomous machine can operate successfully in a diversity of situations without resort to interventien by "higher level" processes, for example, humans. Physical machines are ultimately force or torque controlled dynamical systems: the specification of input torques, whether via syntactic prescriptions or feedback controllers, results in certain classes of vector fields. Control procedures whose resulting vector fields have globally attracting god states may properly be said to evince autonomous behavior. This paper reviews various procedures developed within the Yale Robotics Lab that result in provably autonomous behavior according to the criterion developed above. Simulation results and physical experimental studies suggest the practicability of these methods.