Prof. Dr.-Ing. Helmut Hoyer. Universitíatsstraçe Hagen.

Transcription

1 Control Systems Engineering Group Prof. Dr.-Ing. Helmut Hoyer Universitíatsstraçe Hagen Phone: +49-è0è Fax: +49-è0è Academic staæ: Phone: Dipl.-Ing. Andreas Bischoæ Dr.-Ing. Ulrich Borgolte Dr.-Ing. Michael Gerke Dipl.-Ing. Ivan Ivanov Dr.-Ing. Andreas Jochheim èup to è Dipl.-Ing. Dimitrios Lepentsiotis Dipl.-Ing. Ivan Masar Dipl.-Ing. Christof Ríohrig -1102

2 A "Multiuser Virtual Reality Environment" for communication, interaction and collaboration Andreas Bischoæ To provide communication facilities to a group of students, typical synchronous communication techniques like videoconferencing are not suitable because of bandwidth limitations. A videoconference with more than two communication partners is a typical point-tomultipoint application. If a true collaboration of all partners is desired, the partner with the smallest bandwidth limits the communication. Applications like teleoperated laboratories, seminars or a virtual cafeteria, require a real interaction between students and the tutor, so a bandwidth-saving way ofinteraction is required as an alternative to the video Figure 1: Multiuser VR-Environment based communication. Pure text-based communication èchatè does not meet our requirements because a multiuser teleoperated laboratory application needs the possibility of real interaction. The tutor has to be enabled to introduce and to explain the details of the experiment by some kind of visual representation of the experiment. In this collaborative virtual environment only one student at a time has active access to the experiment. In a 3D-chat an avatar tries to mimic the behaviour of the user in virtual reality. An avatar is the representation of a real person in virtual reality. Participants can see other users as in a real world scenario. It is possible to provide the user avatars with some simple gestures to visualize for instance agreement or disagreement. 3D-chats have lesser bandwidth requirements than video-conferences, because only events are transmitted. The realized collaborative environment ë1ë is divided into two main modules: a rendering and graphics part on the client side and a communication middleware on the server side. On the client side, VRML ë2ë is used to display the virtual 3D environment. VRML as a text-based language is a powerful, nevertheless simple language to build virtual worlds, which include 3D objects, light sources and animations. VRML speciæes an External Authoring Interface èeaiè which can be used by external applications to monitor and control the VRML environment. This is used to update the virtual world with data of the experiment and positions of the other user avatars. The communication middleware is based on the open-source VRML-Multiuser-Software VNET ë3ë which implements its functionality by Java-VRML coupling via the EAI. ë1ë umgebungè ë2ë ë3ë

3 Web-based Teleoperation of Laboratory Experiments Christof Ríohrig, Andreas Bischoæ Within the project ëreal Systems in the virtual Lab" three teleoperated laboratory experiments were developed ë1ë. In 2000, these experiments were used by our students for the ærst time. Students control the experiments exclusively with their standard Web browser, no additional software is needed. A live video stream for viewing the experiment and an optional audio stream helps to provide a laboratory feeling. In 2001 additional features like VRML-visualisation, Augmented Reality and a Multiuser Virtual Reality interface to the experimentation plant were added. The experiment is now useable by a group of students at a given time, all users are represented by so called avatars èrepresentation of a real person in virtual realityè. A real collaboration like in local experimentation is possible. Remote users regard the virtual representation èthe avatarè of the experiment èavehicleè in their VRMLbrowser window. An alternative way of displaying 3D-information of user-avatars is a combination of video and 3D-data, the so called Augmented Reality. In this case a VRML-representation of all the present avatars in a corresponding view to the real camera-system is generated on the server-side and mixed into the live-videostream by a genlock device èa video mixing deviceè. All those virtual and video representations are provided to give the user the impression of being physically present at the experiment, and to allow real collaboration of a group of students. The remote experimentation without the live-video-stream, but including the virtual reality collaborative environment, is still usable under worst case conditions, like a 9600 Baud GSMmobile-phone-connection. Figure 1: Integrated experimentation environment ë1ë ë2ë H. Hoyer, A. Jochheim, C. Ríohrig, A. Bischoæ: Multiuser Environment for a Teleoperated Laboratory, 1st IFAC Conference on Telematics Applications in Automation and Robotics èta2001è, Weingarten, July ë3ë C. Ríohrig, A. Bischoæ: Remote Experimentation in a Collaborative Virtual Environment, ICDE 2001, Díusseldorf, April ë4ë A. Bischoæ, C. Ríohrig: Reale Systeme im virtuellen Labor, kevih-workshop Referenzmodelle netzbasierten Lehrens und Lernens - virtuelle Komponenten der Príasenzlehre, Tíubingen, December ë5ë A. Bischoæ, C. Ríohrig: Multiuser Environment for Remote Experimentation in Control Education, Internet Based Control Education 2001 èbce'01è, Madrid, December 2001.

4 Path Planning for Autonomous Mobile Robots Michael Gerke Long-range route planning is an important precondition to prepare missions for autonomous mobile robots. Usually a lack of energy resources has to be considered, when an exploration path for the mobile robot is predeæned oæ-line. This leads to a path optimization problem with respect to known landscape features. The hybrid approach described here is based on a combination of visibility graphs and genetic algorithms. Visibility graphs enforce the search of feasible paths between start and goal position, whereas genetic algorithms optimize the chain of path segments. Assuming a certain knowledge about the landscape to be traversed by the mobile robot, path planning can be applied to the environmental map. In the case of unstructured outdoor terrain, object conægurations are simple and it is easy to ænd feasible paths based on genetic search only. Chromosome paths are calculated based on this map. After some hundred generations of evolution ècrossover, mutation, reproduction, tighteningè and permanent selection of ættest chromosomes, the optimal path is found between two given locations of the robots mission. But in the case of indoor environments many diæerent restrictions occur. For example, path planning within a meander-shaped free-space conæguration èe.g. a staircaseè is hard to solve with any random search algorithm. Here the search of feasible paths has to be carried out with a systematic approach such as visibility graphs. The example presented here gives an impression of hybrid path planning results. ë 1 ë ë 2 ë M. Gemeinder, M. Gerke: GA-based Search for Paths with Minimal Energy Consumption for Mobile Robot Systems, Int. Conf. on Computational Intelligence, Dortmund, October R. Carduck: Optimization of a Genetic Path Planner for Mobile Robots, Diploma Thesis, University of Hagen, August 2001.

5 The 'LearNet' project: On-line Experimentation via Internet Ivan Ivanov and Ivan Masar We report here of the 'LearNet' project which is funded by the BMBF ègerman Federal Ministry of Education and Researchè. The purpose of this project is to provide a collection of on-line experiments for control theory and practice at several German universities. The project's runtime is from 03è2001 to 12è2003. At the end of the project, on-line experiments are accessible via Internet to support learners and students all over Germany èand beyondè. Remote laboratory experimentation and on-line courseware is developed, which is highly interactive and enhanced with multimedia animations. Within the project 'LearNet', our group provides a container crane and an inverted pendulum as remote lab experiments. Application of control theory with respect to this particular technical system gives a deeper insight into frequency-domain and time-domain control methods. Fuzzy methods are also considered. Figure 1: Hardware of the remotely controllable experiment ë 1 ë ë 2 ë C. Ríohrig, A. Jochheim: Java-based Framework for Remote Access to Laboratory experiments, IFACèIEEE Symposium on Advances in Control Education, ACE 2000, Gold Coast, Australia, December A. Bischoæ, C. Ríohrig: A Multiuser Environment for Remote Experimentation in Control Education IFAC Workshop on Internet Based Control Education, IBCE'01, Madrid, December 2001.

6 Motion Control of Linear Permanent Magnet Motors with Force Ripple Compensation Christof Ríohrig, Andreas Jochheim Linear motors are beginning to ænd widespread industrial applications, particularly for tasks requiring a high precision in positioning such as various semiconductor fabrication and inspection processes. The main beneæts of linear motors are the high force density achievable and the high positioning precision and accuracy associated with the mechanical simplicity of such systems. The main problem in improving the tracking performance of linear permanent magnet motors is the presence of force ripple caused by the irregular magnetic æeld of the permanent magnets and inaccuracy in electronic commutation by the servo ampliæer. Compensation of the force ripple is applied with a model based approach, because force ripple is a highly reproducible and time-invariant disturbance. The force ripple model is based on a Fourier series approximation and is identiæed by measuring the control signal in a closed position control loop for diæerent load forces. No additional sensors èe.g. force sensorsè are necessary. Feedforward C ontro ler α(x) Reference Position x ref e Position Feedback C ontro ler u fb u ff u' β( x) C ontrol Signalu Servo Am plifier Current i Linear Motor Position x Figure 1: Compensation of Force Ripple The model considers both, current-dependent and current-independent force ripple. The compensation of the force ripple is applied with input-output linearization of the controller output u 0 and the servo ampliæer input u. The function æèxè compensates the eæect of the current-dependent ripple, the function æèx è compensates the eæect of currentindependent ripple. u èx; u 0 è=æèxè+æèxè u 0 X è 2çk! X è 2çk æ èxè =k c x + a k sin èx + c k è ; æ èxè =1+ b k sin k2m æ ç p k2m æ ç p! èx + d k è ë1ë C. Ríohrig, A. Jochheim: Identiæcation and Compensation of Force Ripple in Linear Permanent Magnet Motors, 2001 American Control Conference èacc 2001è, Arlington, Virginia, June ë2ë C. Ríohrig, A. Jochheim: Motion Control of Linear Permanent Magnet Motors with Force Ripple Compensation, The Third International Symposium on Linear Drives for Industry Applications èldia 2001è, Nagano, October 2001.

7 Identiæcation of Force Ripple and Friction in Linear Permanent Magnet Motors Jochen Mickel, Christof Ríohrig The main problem in improving the tracking performance of linear permanent magnet motors is the presence of force ripple. One way to overcome this problem is model based compensation of force ripple, which requires the estimation of the respective parameters. Due to the non-linear nature of force ripples a classical linear identiæcation approach is not suitable. Identiæcation by manual force measurements with several diæerent loads at the other hand is time-consuming and ineæective. The main idea of the developed method is to identify the force ripple in a closed position control loop by applying dedicated test vectors to the system, measuring the control signal and positions over time and analyzing the diæerence between the resulting data and a linear system. The method consists of several steps starting with a stiction measurement and a currentindependent force ripple identiæcation. Next a close ætting linear system is identiæed. Finally a current-dependent force ripple identiæcation takes place. Accelerated movements resulting in high control currents help to identify the current-dependent force u x Control Signal Estimated Control Signal Figure 1: Comparison of Control Signals ripples. The output of the method is an estimation of the coeæcients of è1è. Neither additional force sensors nor devices for position adjustment are necessary. The position of the carriage is obtained from an incremental linear optical encoder. In order to avoid inaccuracy by stiction the measurement is achieved with moving carriage. F s èx; uè = F f 0 + k f x, F c0 sgn è _xè+f k0 sinè! k0 x + ' k0 è+ è1è +u èf k1 sinè! k1 x + ' k1 è+f k2 sinè! k2 x + ' k2 èè : The ægure compares the estimated control signal with the measured control signal. The diæerences between the curves are caused by inaccuracies in position and size of individual magnets. The model assumes perfect periodicity of the ripple force. ë1ë J. Mickel: Identiækation nichtlinearer Systemanteile bei elektrischen Antrieben, Diplomarbeit, FernUniversitíat Hagen, 2001.