The juggler model consists of two mechanical vertical arms that are used for juggling as in the case of a human juggler. Each of these two arms is composed of a carriage, moving in the vertical direction and carrying a smaller motor. The actual arm is mounted on the motor and is able to move horizontally. Thus the arms are able to toss over a ball between each other on a parabolical trajectory. A high-speed camera is placed in front of the juggler to detect the movement of individual balls and to close a visual feedback of the actual ball position. Kalman filter has been used to predict the catching point of the ball. The connection to the control system has been realised using the TCP/IP suite because of the camera drivers, which were available just for the Windows operating system without any real-time capabilities. Although the camera frame rate is high enough to supply current ball position into the control loop, due to the nature of the TCP/IP communication and the Windows non-preemptive scheduling, the Kalman filtering was necessary to be used.
The above described model of the juggler was constructed and realised within two master theses of Jan Pšenička and Tomáš Němec. However, additional improvements are needed to be able to accomplish the primary aim of this model to be used for distant learning over the Internet and to make the up-to-date motion-control technologies accessible to the students as well as other interested people from the industry. Such improvements are focused in this proposed project.
The requirement for a model accessible using the distant access is to be constructed in such a way to be able to be operated and manipulated without presence of any person. All operation must be able to be performed remotely. Therefore, another vertical module is going to be added to pick up fallen balls from the floor. Such a module is supposed to get a fallen ball and throw it up to be caught by one of the two main arm used for juggling. As soon as a ball is in the air, it can be tossed over by the two main arms. The high-speed camera is supposed to enable robust control via the visual feedback by improving the original prediction algorithm of the Kalman filter.
The performance of the motors and servo drives has been optimised in order for them to utilise their power as much as possible. The juggling with three billiard balls puts high demands on the servo drives and motors and they are supposed to operate on the edge of their capabilities. This puts also high requirements on the dynamic performance of the control system. Rules are going to be created that can also be used for other systems in their design phase in order to optimise the drive power utilisation and thus to avoid drive overloading and overheating.
To accomplish the optimisation task, a mathematical model of the juggler is going to be created. Consequently, the cam profile parameters of the individual axes can be computed rigorously. PLCopen libraries are going to be used to manage individual cam profiles.
The model of the juggler is based on the B&R architecture. An important and feature of this architecture, which is worth mentioning, is that it provides an all-in-one solution covering all requirements posed on each particular unit of the system, and thus offering a profitable performance ratio. All particular units except PC with machine vision software are connected using the Ethernet Powerlink. The PC machine software is connected using the PVI protocol.
Robot has skills!
