Insertions Using Geometric Analysis and Hybrid Force-Position Control on a PUMA 560 with VAL II

Automatic programming of insertions is an essential step in achieving a truly flexible manufacturing environment. We present techniques based on active compliance implemented with hybrid force-position control capable of inserting a wide variety of shaped pegs. These techniques provide a significant step towards an automatically programmed flexible manufacturing environment. It will be necessary to reduce the programming difficulty of key tasks before robots can be conveniently used to perform assembly operations in truly flexible manufacturing operations. One of these critical operations is insertion, exemplified by the familiar “peg-in-the-hole” problem. Much has been written about solving the case of a chamfered round peg in a round hole. Little is known about solving this problem for more complex shapes, let alone threaded or bayonet insertions. In our work we have developed a general approach to oriented insertions that uses geometric properties of the object to control the behavior of a hybrid force-position controlled robot. Mason introduced a model for position and force control for manipulators . In this model the degrees of freedom of a manipulator are partitioned into orthogonal subspaces representing the force controlled and the position controlled motions of the manipulator. This model provides a concise means of describing complex tasks, although in some cases the description is difficult to interpret. Raibert and Craig implemented a controller based on Mason’s model and performed some experiments within the capability of a two degree of freedom manipulator [Raibert and Craig, 19811. In our work we have developed a means of hybrid force-position control for a PUMA 560 using the VAL II controller. Our technique allows six dimensional subspace partitioning into force and position controlled subspaces. The current implementation is restricted to subspace components being associated with the Cartesian axes of the tool frame. Our implementation extends Mason’s model in that it provides a “guarded move” [Will, 19751 capability for both force and position constrained movements. In this paper we describe how we implemented hybrid force-position control and how we applied it to performing force-directed oriented insertions based on geometric constraints. The relevant portion of the Sandia Intelligent Robotic Assembly System (SIRAS) is comprised of a PUMA 560 six degree of freedom manipulator equipped with an Astek (now Barry Wright Corp.) FS6-120A 6-axis force-torque sensing wrist, an unmodified Unimation VAL II controller, a PDP 11/73 arm monitor, and a DEC microVax II task control computer. All user interaction is through the microVax in SCHEME, a dialect of LISP. The microVax communicates with the PDP 11/73 monitor which handles all communications to and from the Unimation controller. The arm monitor also provides the interface between the force sensing wrist and the Unimation controller. The VAL II language includes an ALTER mode in which the controller polls the ALTER port every 28 ms (the basic timing cycle of the controller) for a set of translational and rotational offsets for the tool frame from the nominal position dictated by the current movement command. This mode continues until an END ALTER command is received. These offsets can be either cumulative or not, causing the manipulator to act as a dashpot or a spring, respectively. Our approach to hybrid force-position control was to implement a program on the arm monitor that calls the ALTER program on the Unimation controller and provides cumulative offsets to the ALTER port based on readings from the force sensing wrist and the parameters from the SCHEME command. The format of the SCHEME command is (MCOMPLY GAIN BIAS THRESHOLD