A Review on the Dynamic Control of Parallel Kinematic Machines: Theory and Experiments

Abstract

In this article, we review the dynamic control of parallel kinematic machines. It is shown that the classical control strategies from serial robotics generally used for parallel kinematic machine have to be rethought. Indeed, it is first shown that the joint space control is not relevant for these mechanisms for several reasons such as mechanical behavior or computational efficiency. Consequently, Cartesian space control should be preferred over joint space control. Nevertheless, some modifications to the well-known Cartesian space control strategies of serial robotics are proposed to make them perfectly suited to parallel kinematic machines, particularly a solution using an exteroceptive measure of the end-effector pose. The expected improvement in terms of accuracy, stability and robustness are discussed. A comparison between the main presented strategies is finally performed both in simulation and experiments.

Linear dynamic modeling of parallel kinematic manipulators from observable kinematic elements

“…Since the inverse kinematic problem is usually well posed for a parallel robot, a more efficient choice is to use the end-effector Cartesian pose and its velocity as dynamic state variables [20]. This choice leads to efficient models as long as one is able to measure or estimate the pose and velocity of the end-effector in the Cartesian space.…”

“…This choice leads to efficient models as long as one is able to measure or estimate the pose and velocity of the end-effector in the Cartesian space. For instance, it was shown that such a choice makes Khalil's methodology totally linear since it replaces the non-linear computation of the forward kinematics of each leg [20]. However, there remains another loss of efficiency: one can not directly measure in the Cartesian space, so one has to estimate such variables, either through a mechanism or by optical means (e.g., laser, vision).…”

Robust Modeling and Control Issues of Parallel Manipulators with Actuation Redundancy

“…Now the feedforward term requires the inverse dynamics solution (11). These control methods, originally derived for non-redundantly actuated systems, can be directly adopted for RA-PKM as in Cheng et al (2003); Müller (2005); Paccot et al (2009). The APD can be used in the form (omitting null-space components)…”

“…In other words, it is assumed that all actuators can be independently controlled without mutual interference. This applies to serial manipulators as well as to PKM without actuation redundancy as summarized in Paccot et al (2009) andThanh et al (2009). However, since in case of RA-PKM more actuators are activated than required, decentralized control of RA-PKM naturally leads to conflicting control forces, reflected by undesired prestresses and an increased power consumption as observed in Saglia et al (2009),Valasek et al (2005.…”

Mechatronic Development and Dynamic Control of a 3-DOF Parallel Manipulator

“…The reason for building an open architecture is that the control is a field where there is still great potential for study in order to improve its accuracy (Paccot et al, 2009). Thus, the robot developed can be used as a control scheme test bench.…”