Controlling a contactless planar actuator with manipulator

An existing magnetically levitated planar actuator with manipulator has been studied and improved from a control point of view. This prototype consists of a magnetically levitated six-degree-of-freedom (6-DOF) planar actuator with moving magnets, with a 2-DOF manipulator on top of it. This system contains three different contactless technologies: contactless bearing and propulsion of the planar actuator, wireless powering of the manipulator, and wireless communication and control of the manipulator. The planar actuator (PA) consists of a Halbach magnet array, which is levitated and controlled in all six DOF’s above a stationary coil array. The PA is propelled in two horizontal translational DOF’s while the other four DOF’s are stabilized to accomplish a stiff bearing. Each active coil contributes to the production of forces and torques acting on the magnet array. Since the number of active coils is much larger than the number of DOF’s, the desired force production can be distributed over many coils. Therefore, a commutation algorithm has to be used to invert the mapping of the forces and torques exerted by the set of active coils as a function of the coil currents and the position and orientation of the translator. One method for linearization and decoupling of the forces and torques was developed in the past. The method is called direct wrench decoupling and guaranties minimal dissipation of energy. However, no constraints on the maximum current can be given. This study proposes two novel, norm-based commutation methods: l8-norm and clipped l2-norm based commutation. Both methods can put bounds on the maximum currents in the coils to prevent saturation of the current amplifiers. The first method focuses on minimization of the maximum current whereas the second method limits the peak current while it minimizes the power losses. Consequently, a higher acceleration of the translator can be achieved and/or less powerful (cheaper) current amplifiers can be utilized and/or fewer commutation errors arise. Only a long-stroke translational movement of the moving magnet planar actuators has been considered in the past. The possibility of a completely propelled and controlled rotation about the vertical axis instead of just stabilizing it for bearing has been analyzed in this thesis from a control point of view. Enhancing the planar actuator with a long-range rotation will increase its utility value and opens new application areas. Based on this investigation, a novel coil array with a triangular grid of rounded coils has been proposed for better controllability in any orientation of the PA. In addition, other coil and magnet topologies have been studied from a control point of view for their suitability for full rotation. The influence of different kinds of error-causes on the commutation precision has been studied. From this investigation, it has been found that the offsets of the measurement system have the highest influence on the precision of the commutation. Investigation of the convergence of the procedure for estimation and elimination of these offsets has been performed. Although it was not proven that the procedure could be applied on the whole workspace of the PA, the convergence has been shown at least for all the investigated points. From this investigation, convergence for any position in the workspace of the PA is expected. It was found that it is possible to use the procedure also with different topologies and with different commutations. A novel wireless link has been developed for the real-time control of a fast motion system. The wireless link communicates via infrared-light transceivers and the link has a delay and a packet-loss ratio almost indistinguishable from the wired connection for the bandwidth of the system up to several kilohertz. The clipped l2-norm based commutation method has been successfully tested on the experimental setup after improving the measurement system, the contactless energy transfer and the wireless communication. With a new, interferometer sensor system, a well-controlled PA with two long-stroke DOF’s has become available. Improved contactless energy transfer does not cause increased electromagnetic interference during switching between the primary coils any more and the wireless connection using the infrared link provides a reliable communication channel between the manipulator and the fixed world. Several control approaches have been tested on the experimental setup. Both, the classical PID control, Sliding-mode control and Iterative learning control have been implemented. Each controller brought better performance than the previous one. Also, a fourth-order trajectory and enhanced feedforward control helped to improve performance. Finally, the tracking errors, in comparison to the initial situation, were reduced by a factor 10 (and even more than by a factor 50 with deactivated contactless energy transfer) while the velocity and acceleration of the system were a factor 4 and 14, respectively, higher