An Instant Messaging Framework for the Virtual University
Andreas Bischoff
A modern virtual university environment requires communica-
Figure 1: JabberClient MirandaIM
tion tools for synchronous events like online workshops and onlinepractice. Computer Supported Collaborative Learning (CSCL)environments for synchronous events are still difficult to handle.For synchronous communication like audio- or video-conferencing,open TCP- or UDP-ports are required. Especially upcoming lim-itations of Internet access like firewalls and NAT-routers increasethe need of backup solutions for synchronous events.
To provide communication tools for student workgroups and asa backup strategy for synchronous communication we use Jab-ber [1] based Instant Messaging (IM) Server and Clients. InstantMessaging is a very reliable tool to support the users. If technicaldifficulties arise, in the case of modem-users, the only communica-tion channel is occupied by the Internet connection. Particularlywith regard to user awareness the “presence”-feature of InstantMessaging Clients is very important and useful. If a user loses theconnection to the Internet during a synchronous event the tutor isinformed in real-time. Jabber itself is an XML-based open sourceprotocol for Instant Messaging services. Jabber-based software is used by over a millionusers worldwide. The protocol is maintained by the Jabber Software Foundation.
Advantages of Jabber over conventional commercial IM services are interoperability withexisting IM-services, a XML-based open source protocol and an easy way to adapt theserver to existing services. Robust and secure clients for nearly all operating systemand mobile devices like PDA’s and cellular phones are available. We have build up aJabber based Instant Messaging service which is interoperable with the virtual universityenvironment of the University of Hagen [2]. We use the C++ based open source ‘Jabberd’with an additional xdb auth/check module [3] to provide a connection to the existingLDAP-directory service at the University of Hagen. This modification is very convenientfor users because no extra passwords and administrational effort is necessary [4].
the IEEE 802.11b standard are wellknown and widely used to support LANinfrastructures [1] . Our new approachis to use existing WLAN accesspoints forlocalisation of PDA and notebook usersor autonomous mobile devices. Localisa-tion is useful to provide location basedservices, similar to cellular phone appli-cations. The tracking of mobile WLANdevices can be used to support the lo-calisation of mobile robots. The advan-tage of this approach is to use the neces-sary communication medium for locali-sation also. All modern wireless LANcard drivers provide information about received beacon signals and the field strength ofthe available accesspoints. The opensource hostap (host accesspoint) Linux-kernel driverfor Intersil-Prism II wireless LAN chipsets is also enabled to receive this information.With a capable calibration and a minimum of 3 received accesspoints, a localisation withan average precision of 1 meter is possible. This precision is adequate for location basedservices. For the localisation of mobile robots this precision is sufficient for a roughestimation of the position and could be used as part of a sensor-fusion.
Our first implementation of a localisation system
Figure 2: ‘mixed reality’ VRML envi-ronment with located WLAN-user
uses the hostap driver together with the Linuxwireless-tools [2] and own extensions for a roughestimation of the positions. A more accurate cali-bration is possible with a new commercial applica-tion, the Ekahau positioning engine [3]. This soft-ware provides Java interface classes which can eas-ily be connected with a representation of the userin a virtual reality environment [4]. In this con-text localisation in wireless LANs can be used ascheap and robust tracking system. Tracking sys-tems for areas of around 3m3 are usually very ex-pensive (about 10,000 Euros), so this solution is acheap alternative. A sample calibration for the of-
fice area of the Control Systems Engineering group (PRT) and a tracked path is shownin figure 1. With this kind of tracking system a “mixed reality” environment, a sharedspace for collaboration of local and distant user can be realised (figure 2).
Within the EU’s 5th framework programme, the Network of Excellence “EURON”(European Robotics Research Network) was set up [1]. The objective was to imple-ment and maintain a network of excellence that allows coordination of research, teachingand education, academic-industry collaboration, and publications and conferences in thearea of robotics, to facilitate addressing of issues of interest to institutions and companiesthroughout Europe. The aim was to provide a fruitful basis that allows Europe to remainat the forefront of robotics both in terms of research and industrial products.
EURON will finish its work in 2004. To continue the successful network, a proposal forEURON II was submitted for the 6th framework programme. This proposal has beenaccepted by the European Commission and will commence right after EURON in 2004.Our research team will take part in EURON II.
EURON II is situated in the EU-IST-FET initiative “Beyond Robotics” (BR). The areaof robotics is expanding into a number of new application domains and it offers newopportunities to study human-robot augmentation, advanced autonomy and user inter-faces, and in addition integration of communities of “simple” robots. This opens newopportunities in terms of task achievement and flexibility. All these areas are addressedwithin the BR initiative. The network focuses on four primary activities:
• As the Commission of the European Communities has launched a number of inte-grated projects (IPs) in BR there is naturally a need for coordination across theseprojects, and there is at the same time a need to provide a long-term roadmap forthe initiative to consider its integration in the long-term goals of the commission(and the involved communities). Research coordination will provide the mecha-nisms for overall integration of the research within BR.
• A number of emerging ideas originating from the IPs and the communities in gen-eral might warrant further consideration to determine their feasibility and potentialintegration into the general initiative. To ensure this, the network has a joint re-search programme through which prospective research projects will be sponsored.In addition topical studies across the initiative will be sponsored to allow for close in-tegration of the involved communities into the efforts within the IPs. For specialisedstudies of well defined topics the network might also setup “research ateliers” forsingle venue time limited studies.
• An important part of the BR initiative is the dissemination of results and excellencebeyond the IPs. An effective mechanism for community setup and disseminationof knowledge is through summer schools, standardised educational efforts and newplatforms for teaching. Consequently the network will organise a wide variety ofeducational efforts to provide the required human resources.
• In addition to education there is also a need to build strong ties to existing andnew industries to ensure a long-term pick up of the results from the initiative. Thisis ensured through a concerted effort related to establishing an industrial club forindustries that may utilise the generated knowledge.
[1] http://www.euron.org
Robot Soccer
Michael Gerke
Robot Soccer has become an international standard for testing and evaluation of dis-tributed technical multiagent systems. It is distinguished between several robot soccerleagues, starting with a simulated league for software soccer agents, via small size andmiddle size real robots towards legged systems. Simulated soccer teams meet each othervia computer networks, where a soccer server manages the clients (individual players)and provides a graphical simulator for visualization.During the annual RoboCup event (combined with a related scientific conference) theworld championship for all leagues is carried out. The declared goal of the joint effortsis to finally create a humanoid robot soccer team which can win a soccer match againstthe human world champion team according to FIFA rules. The year 2050 is regarded asa feasible deadline for this project.Our research group is involved in the development of a soccer team for the simulatedleague. The purpose of this involvement is twofold:
• Scientific aspects of multirobot cooperation can be tested and evaluated in a highlymotivating competitional situation. Results are projected onto real applications inour robotic labs.
• Students are integrated into our soccer activities during their projects and seminars.Here they learn aspects of multiagent systems and distributed client cooperationas well as object-oriented programming.
Figure 1: Visualization of a simulated soccer match
[1] A. Zolgert: Aufbau eines Soccer Servers unter Linux, Diploma Thesis (in German), Hagen2003
[2] U. Zauner: Einsatz von Neuronales Netzen in verteilten Multi-Agenten-systemen,Diploma Thesis (in German), Hagen 2003
The Blimp Project - New Developments
Michael Gerke
Based on the ‘push’ given by the ‘3rd Mission’ exhibition in the state government buildingof Northrhine-Westfalia in Duesseldorf, November 2002, our research group has focusedon a new area of robotics: autonomously flying robots. We started in 2002 with a remotelycontrolled blimp (airship) which travels in the airspace of our robotics lab. This flightexperiment can be radio controlled from a computer system with the help of an on-boardvideo camera and a web browser including video and control applets. At that point, ahuman operator was still in the control loop.
In the meantime we have acquired a digital signal processor (DSP, Texas Instruments)and some additional sensorics [1] like accelerometers, gyroscopes to implement steady-state control loops thus restricting the aerostatic system from drift due to unexpecteddisturbances.
Figure 1: The blimp as a flying robot
We now proceed with our development to allow trajectory tracking and path following forthe blimp based on visual servoing and gyro servoing. Furthermore we include aspects ofteleoperation to give Internet access to the airship. Access will be provided in the samemanner as in various other remote experiments developed by our research group [2].
[1] M. Gerke: Roboter erobern die dritte Dimension, Jahrbuch 2003 (in German),Gesellschaft der Freunde der FernUniversitat e.V.
[2] H. Hoyer, M.Gerke, I.Masar, I.Ivanov, C.Roehrig, A.Bischoff: Virtual laboratory for Real-Time Control of Inverted Pendulum/Gantry Crane, 11th Mediterranean Conference onControl and Automation (MED’03), Rhodes, Greece 2003
Real Automation Systems in a Virtual Laboratory
Dimitrios Lepentsiotis
Globalisation and short innovation cycles are fostering an explorative information over-flow, demanding corresponding reforms in education and further education. In industryand respective publications, this new paradigm emerges by trendy terms such as “just intime learning”, “training on demand”, or “blended learning”. The main problem withthis is the proper mixture of teaching and learning methods, as well as flexibility in placeand time. New information and communication technologies are the potential base forthese new methods.
To accept the challenge, the Control Systems Engineering Group of the University of Ha-gen and the automation department of Siemens AG joined some nine years ago. Siemensrepresents competency in Programmable Logic Control (PLC), University of Hagen con-tributes well-tried teaching and learning concepts. Both aimed at developing and offeringpractical courses of high quality [1].
There is an increasing impact of the internet as a teaching and learning medium. In thiscontext, the University of Hagen developed the concept of “Virtual University LearningSpace”. The Control Systems Engineering Group contributed the “Real Systems in aVirtual Laboratory”, where students can make control related experiments independentlyin space and time [2]. Now, a new experiment for the area of further vocational educationin PLC has been added [3]. The principle of this experiment is illustrated in figure 1.
Figure 1: Interactive online PLC experiment
The students can program a laboratory based PLC from their PC at home using a con-ventional web browser. The programming environment is the same as used in industrialenvironments. In addition to programming, the basic software package allows modifica-tion, documentation, operation, and supervision. The remote control tool VNC (VirtualNetwork Computing) allows access to the original programming environment which islocated on a NT-server in the laboratory. The experiment can be visualised at home bya video connection with the laboratory. The video connection is powered by Java MediaFramework (JMF). The PLC controls a real model railway with sensors and actors.
[1] prt.fernuni-hagen.de/sps/spsrail/
[2] prt.fernuni-hagen.de/rsvl
[3] prt.fernuni-hagen.de/sps
Using of Artificial Intelligence and Soft Computing Methodsfor Autonomous Mobile Robot Control
Ivan Masar
One of the most important features of nowadays-mobile robots is autonomy during exe-cution of given tasks. The robots must adapt and optimise their movement to fulfil theirmission and learn how to do it better next time. Methods of artificial intelligence andsoft computing help to accomplish this goal.
We present an example of using these methods for control of an autonomous mobilerobot. It is a self-learning neuro-fuzzy controller for trajectory following. This trajectoryis represented in our case by a line drawn on the floor. In industrial practice, it is oftenrealised by wiring built in the floor. As the robot can be used in partially known/unknownenvironment, the coordinates of the trajectory could be generated in advance or on-lineby a path-planner.
The tuning of controller parameters takes place in two steps. In the first phase, thedynamical model of the mobile robot following the line is learned. This model is imple-mented using a multi-layer neural network and it is required for training of the neuro-fuzzycontroller. For the purpose of neural network training, sufficient amount of training data(coordinates of the trajectory, robot position and speed, etc.) must be recorded. Thisdata is recorded either by manual steering of the mobile robot along the desired path, orby using a simple PD controller, which is able to steer the robot at least during low for-ward speed. Eventually, the simulation of an analytically derived dynamic model of themobile robot [2] can be used for acquisition of training data sets. As adaptation methodfor the forward neural model, hybrid learning was chosen because of its fast convergence.This type of learning combines an adaptation of non-linear parameters of the system bya Levenberg-Marquardt algorithm, while the estimation of its linear parameters uses theLSE method.
In a second step, the parameters
)(keL
z-1
Robot Data/Simulation
Neuro-FuzzyController
)(kwL
Neural Network(Forward Model)
z-1
z-1
)(~ keL
)(keC
LearningAlgorithm
ω (k)θ (k)
Figure 1: Control loop with neuro-fuzzy controller
of the neuro-fuzzy controller areadapted. This process takes placeduring trial trajectory following. Itcan be realized again using simula-tions, or by experiments with a realrobot. In both cases, several quali-tative measures like sum of squarederrors, speed of line following, max-imal overshoot, etc. are evaluatedin periodical intervals and used forthe controller adaptation.
Next developments will be focused on speeding-up of on-line learning and optimalizationof robot movement and steering.
[1] I. Masar, M. Gerke: Mobile Roboter im Wettkampf, Proc. 13. Workshop Fuzzy Systems,FZKA-6900, Dortmund, November 2003
[2] Patrick F. Muir: Modeling and Control of wheeled mobile robots, Ph.D. Thesis, UMI,Pittsburgh, 1988
The LearnNet Project
Ivan Masar, Michael Gerke, Andreas Bischoff, Christof Rohrig
After three years of development, the BMBF project ”LearNet” will finish this year.After succesful evaluation of the common platform for learning and experimentation viaInternet, the network of on-line experiments and servers with various courses will beavailable to the students from all over Germany and the world to support them to studycontrol theory and process automation. Our contributions to this project are:
• Design and implementation of a virtual laboratory for on-line controlof the inverted pendulum/gantry crane system - For that purpose, a novelarchitecture of the virtual lab was developed. Several new components, includingxPC Target Server for real-time operation of the experiment and communicationwith the client, a Control Applet for interfacing to the lab, Multimedia Reflector forvideo/sound streaming and an Analyser Applet for evaluation of the experimentaldata, provide a basic framework for future labs.
• Development of the reservation system - By means of this database system,the students can schedule the date and the length of their experiments by them-selves.
• Generation of several on-line accessible course materials (fuzzy control,state space control, etc.)
ExperimentSet-Up andOperation
MultimediaStream
Virtual Reality
Evaluation of theExperimental Data
Interface to theReservation System
Figure 1: Interface to the Virtual Lab with Analyser and Reservation sytem
[1] http://www.learnet.de
[2] H. Hoyer, M. Gerke, Ch. Rohrig, A. Bischoff, I. Masar and I. Ivanov: Reale Systeme im’virtuellen Labor’/Real Systems in ’Virtual Laboratory’, Automatisierungstechnik (at),Volume 11, 2003
[3] H. Hoyer, M. Gerke, I. Masar, I. Ivanov, C. Rohrig and A. Bischoff: Virtual Laboratoryfor Real-Time Control of Inverted Pendulum/Gantry Crane, 11th IEEE MediterraneanConference on Control and Automation, MED’03, Rhodos, Greece, June 2003
Mobile robot F.A.A.K./Istrobot 2003Ivan Masar and Frank Metzler
Mobile robots go to competitions. Their capabilities grow with increasing miniaturiza-tion of their components, development of new actuators and permanent improvement ofperformance of their control and sensing electronics. The better perception of the en-vironment by sensors, digital signal and image processing, new types of algorithms forautonomous operation, self-learning and reactions on outer stimulus based on methodsof artificial intelligence and soft computing, as well as powerful and ever more powerefficient hardware make various novel comprehensive scenarios feasible.
One of the most innovative impulses in the field of mobile robots is robot competition.New designs are needed due to restrictions on the robot size (and thereby on quantityof portable instrumentation and power supplies); ambition to accomplish the given taskin the shortest possible time with the required precision; and constrained time for devel-opment. Although the goals of the competition are usually not very complicated, it isalways interesting to observe various new or modified solutions, which help to win thecontest.
In the past year, we adapted the mobile robot F.A.A.K. (Figure 1), which was completelydesigned at our department, for the robot contest ISTROBOT 2003. It is the largestmobile robots competition in countries of Eastern Europe and takes place every year inthe capital of Slovakia, Bratislava, under the auspices of the Department of Control ofSlovak Technical University.
Our robot is designed to fulfil the task of two main events - ’Pathfinder’ and ’Mouse inLabyrinth’. A non-conventional kinematic structure with four steered wheels was used,which enables the robot to perform various types of motion, depending on the type ofapplied steering algorithm. Therefore, the robot can change not only the curve radius,but also its centre of rotation ((a) and (b), respectively). Moreover, it can rotate aroundits vertical axis (c) or cruise (d).
All robot axes and joints
�����
ICR
���
�ICR
�����
���
�
(a) (b)
(c) (d)
ICR
Figure 1: F.A.A.K. robot and its types of steering
are actuated by DC mo-tors. On-board, severalsensor types (IR distancesensors, CCD line camera)are installed. Because ofhigh performance and richperipherals, we chose theTMS320LF2407A DSP ofTI to control the mobile ro-bot. Although the robot isfully autonomous, it is alsoequipped with a radio mod-ule for communication with a developing platform, based on the Matlab/Simulink soft-ware package. This interconnection enables on-line visualisation of the robot motion bymeans of its virtual model as well as processing of various data. As the next enhance-ment, a new on-board image processing system using a small digital camera and one ofthe powerful DSPs from TI’s C5000-family is intended.
[1] see www.robotika.sk
Force Ripple Compensation of Linear Synchronous Motors
Christof Rohrig
Linear synchronous motors are beginning to find widespread industrial applications, par-ticularly for tasks requiring a high precision in positioning such as various semiconductorfabrication and inspection processes. The main benefits of linear motors are the highforce density achievable and the high positioning precision and accuracy associated withthe mechanical simplicity of such systems.
The main problem in improving the tracking performance of linear permanent magnetmotors is the presence of force ripple caused by inaccuracy in commutation of the phasecurrents by the servo amplifier. Force ripple is an electro-magnetic effect and causes aperiodic variation of the force. Only if the back-EMF waveforms of the motor phases aresinusoidal, sinusoidal current waveforms generate ideally smooth force.
The proposed compensation method is based on optimized current waveforms whichproduces minimal copper losses and maximize motor efficiency. The method for forceripple compensation consists of three stages. In every stage different harmonic waves ofthe force ripple spectrum are reduced.
1 2 3 4 5 6 7 8 90
2
4
6
8
10
12Without Optimization
Am
plitu
de [%
]
1 2 3 4 5 6 7 8 90
2
4
6
8
10
12Step 1
Am
plitu
de [%
]
1 2 3 4 5 6 7 8 90
2
4
6
8
10
12Step 2
Am
plitu
de [%
]
1 2 3 4 5 6 7 8 90
2
4
6
8
10
12Step 3
Am
plitu
de [%
]
Figure 1: Force ripple spectra
The figure shows the force ripple spectra in the different steps of the optimization process.
[1] C. Rohrig: Zur Lageregelung synchroner Linearmotoren fur hochdynamische Anwendun-gen unter besonderer Berucksichtigung der Kraftwelligkeit, Dissertation, FernUniversitatin Hagen, Fachbereich Elektrotechnik, 2003, VDI-Verlag, Reihe 8, Band 1016, Dusseldorf,2003
