Open-loop Velocity Planning to Mitigate the Stiction Effect in Pushing Positioning

Actuation by pushing has been studied as both an economic and flexible alternative to traditional pick-and-place part positioning. Pushing actuation reduces the need for specialized end effector tooling, and can simplify the underlying automation mechanism (e.g., a 2DOF rotational positioner in place of a 6-DOF robot). However, the interaction of the positioning command with the nonlinear behavior of sliding friction introduces a new source of position error. The stick-slip effect of friction (i.e., stiction) reduces local controllability in precision positioning applications. In this paper, we describe a method for identification of dominant stiction frequency across a range of actuation velocities and compensation of the input command signal to avoid actuating the part at or near the stiction frequency. The result when compared with simple constant-velocity actuation is a reduction in variation of actuating force and an improvement in achievable position accuracy through more consistent control of part sliding at end-of-actuation-stroke. INTRODUCTION Actuation by pushing has been studied as both an economic and flexible alternative to traditional pick-and-place part positioning. Using pushing actuation, a number of system improvements can be realized. Foremost, the complexity of positioning systems can be reduced. In a traditional pick-and-place system, space at least 3 degrees of freedom are typically employed: two for Cartesian positioning and one for changes in elevation. Employing a pushing actuating system allows for movement of a part along its current plane. This allows the part to not only be precisely positioned, but it can also be programmed to follow a predetermined actuation path; for example to clear obstacles. Additionally, pushing actuating can be employed using a single fixed pusher tip rather than the part-dependent tooling commonly found in pickand-place systems. This reduces system complexity cost and improves flexibility in multipart operations. However, actuation by pushing can introduce a stick-slip effect (stiction), which can degrade system accuracy. The system friction and stiction effect can be simply modeled, and system behavior predicted. We propose to use this information to aid in velocity planning for the push actuation, in order to mitigate the stiction effect. ACTUATION BY PUSHING In the past 20 years, there have been numerous research efforts in the field of precision positioning by sliding the target object across a surface. Peshkin and Sanderson describe the motion of a sliding workpiece for all possible pressure distributions on the support surface [Peshkin and Sanderson 1988]. Zesch and Fearing explore force-controlled pushing for microparts with positional results in the 1μm range [Zesch and Fearing 1998]. Lynch and Mason have done extensive work on planning and control for stable pushing in the application of robotic manipulation as an alternative to pickand-place positioning, including feasibility studies through both kinematic and force analyses [Lynch and Mason 1995; Lynch and Mason 1996]. Lynch also explores friction estimation for pushed objects and open-loop control for pushing the general polygonal shape, characterized by the “maneuverability” property [Lynch 1993; Lynch 1999]. A number of research efforts have been directed at positioning parts using impact or single-contactor pushing actuation. Benefits are a more inexpensive and flexible actuation system that can be designed for very large or very small parts. Research in application of impact to positioning has mainly been focused on static initial and end conditions and single impact system input. That is, a part initially at translational and rotational rest is struck once to impart a velocity, and then allowed to come to rest under environmental conditions (typically friction). Application of these concepts to impactbased static positioning systems is treated separately by [Mendes, Nishimura et al. 1996] in the printed circuit board positioner, by [Liu, Higuchi et al. 2003] in their piezoelectric positioning table, as well as by [Siebenhaar 2004] in electromechanical hammer control. FRICTION MODELING APPROACHES Friction is present in all mechanical systems, and contributes significantly to force analysis and control of motion systems. In this case, it is important to fully understand and accurately model friction when developing an idealistic model of the physical system. There exists substantial research on modeling of static and dynamic friction, both in the idealized linear case and the nonlinear case. Both [Olsson, Astrom et al. 1998] and [Canudasde-Wit, Olsson et al. 1995] give a comprehensive overview of the major static and dynamic friction models utilized in practice. These ideas are extended to the special case of low velocity friction compensation by [Adams and Payandeh 1996]. The classical friction model was derived by Coulomb and is of the linearized form