About Robot Laboratory, Lund Institute of Technology

ABB Robot

Research Direction

Several research interests are represented in Robotics Lab:

Robot control systems and other manufacturing equipment are traditionally closed. This circumstance has hampered system integration of manipulators, sensors and other equipment. As a result, such system integration has often been made at an unsuitably high hierarchical level.

The purpose of past and present projects is to show how to organize open robot control systems and to verify these ideas by means of experimental verification.

As a part of this research, we have developed several experimental open robot control systems. The systems are built around industrially available robots that have been reconfigured for experimental purposes.

The developed specific robot interfaces and the integration of the robots into a complete system forms a unique environment for testing and development of algorithms for improvement of performance, sensor integration, programming automation and autonomous operation.

History of Robotics Research at Lund University, Department of Automatic Control

As robots provide tangible demonstrators of  the merits of control for audiences  at all levels of scientific sophistication ranging from kindergarten to students and to faculty members, the Robotics Laboratory of Department of Automatic Control is one of the favorite meeting points for visitors and guests.

Robotics is a research area with intimate relationships with control and computer science, yet with an independent multidisciplinary character. Beside the mechanical  nature of robots, robotics requires all the cybernetic elements of control, communication  and computation and robotics  can never be reduced to a mere application of control theory. Apart from control science, successful robotics require efforts in mechanical design, motor drives or other actuators, sensor technology and software engineering. Therefore, the history of robotics at our department is closely relate to the history of the enabling technologies, including the projects and researchers that contributed to the developments. In this perspective, the progress within digital control, real-time systems, visual feedback, and control of mechanical servos, deserves some attention. The preliminaries of robotics at the Dept Automatic Control have at least a history of 30 years.

Preliminaries (1970-1984)

To the purposes of the educational laboratory, Leif Andersson made DC servo equipment with an interface to analog computers. The DC servos permitted sensor feedback of position and velocity and were used for many years in the basic course for most students. Another important  prelude to robotics was the design of the ball-and-beam process which was designed by Johan Wieslander and Karl J. Åström with mechanical design and construction by Rolf Braun and Hans Libelius, technician at our neighboring Dept Solid Mechanics. Much of this design work was done in the framework of Johan Wieslander's thesis work and the equipment brought early attention to problems of nonlinear control and real-time control. Early control was made by means of computer-based control using a PDP15  from Digital Equipment, Inc.

Computer-based control with all attempts to connect computers to control equipment and other peripheral devices also stimulated networking. Early efforts in network-based control was made by Rolf Syding,  Ulf Borisson, Leif Andersson and colleagues in efforts to control an ore crusher  of LKAB-a mining company in northern Scandinavia. By connecting the ore crushing plant over the telecommunication  network, a geographically long feedback loop (>3000km) was connecting to real-time control  in a PDP15 computer. Rolf Braun designed an actuator inteface

During his time as a graduate student, Lars Nielsen made early efforts in visual servoing, partly in cooperation with Gunnar Sparr, Dept Mathematics, Lund. During his thesis work, Lars Nielsen developed a small, but impressive in terms of its purpose, mobile robotics laboratory. After his postdoc visit to CalTech in 1986, Nielsen returned to Lund and took iniatives to continued robotics research in cooperation with Ola Dahl and Klas Nilsson who had joined Dept. Automatic Control as PhD candidates. The focus now was industrial robotics, which includes most of the challenges within mobile and autonomous robotics, but also brings forward industrial relevance and performance demands. With the new focus, with the PhD students Ola Dahl and Klas Nilsson, and with a great interest from Karl J. Åström, Lars Nielsen, Rolf Johansson and other faculty members, new experimental facilities were created

After his dissertation on multivariable adaptive control in 1983, Rolf Johansson started theoretical work in nonlinear control with an application perspective on mechanical systems. Another source of inspiration came from biological inspiration acquired from his experience in neuroscience research. New algorithms and analytical results for adaptive control and nonlinear optimal control were published in 1987-1990.
 

Early days  of the Robotics Laboratory at Dept. Automatic Control (1983--1992)

Initiated by Karl J. Åström  and with a great interest and support from Lars Nielsen, Rolf Johansson and others at the department, the first industrial manipulator was acquired from ASEA in 1984. At ASEA there was an increasing need for computer-based control engineering tools, which was something that was being developed at the department. When the first industrial robot---an ASEA Irb-6 (designed around 1978) equipped with the  S2 control system designed during 1981-82---was acquired, it was actually made in exchange  of the   control design and simulation software packages (Simnon, Synpac, Modpac, and Polpac) developed and sold by the department. At ASEA, it was the manager of technical development, Hans Skoog that approved the deal proposed by Klas Nilsson. Also, the first PhD student with a clear focus on robot control started, namely Ola Dahl started in 1984. Klas  Nilsson, who had been a student at Lund made his master thesis work

at ABB Robotics, and started his professional career as a control engineer at ABB Robotics. Few  individuals  have greater credit than Klas  in the creation of a Robotics Laboratory. While still at ABB Robotics, Klas was instrumental already in early use of the Irb-6. When beginning as a graduate student at the department in 1988, he started work on hardware modification to permit highly modifiable control software architectures.When beginning as a graduate student  at the department in 1988 with Lars Nielsen as graduate advisor, he started work on hardware modifications to permit highly configurable control software architectures.
Important interfacing work was made in cooperation with Rolf Braun. There was an earlier interface of a pure analog type, permitting some experiments in restricted parts of the working range of the robot. Since Klas knew the inner workings of the system, it was possible to interface external computers with the safety logic of the embedded controller. Important interfacing work was made in cooperation with Rolf Braun who reeplaced all original electronic hardware (except for power electronics).  This work on the Irb-6 started in 1990 and the Irb-2000 was modified in 1992.

During his time as a graduate student under guidance of Ioan Landau (CNRS at ENSIEG-LAG, Grenoble,  France), Carlos Canudas-de-Wit  spent the greater part of 1985 in Lund as a guest of Karl J. Åström. He was  using methods of  system identification and friction modeling to develop friction compensation in servo applications. Interaction with Rolf Johansson,  who at that time was developing system identificstion methods for continuous-time models, combined with Carlos interest in friction modeling and eventually this gave rise to the LuGre friction model.
 

Steve Murphy who was then graduate student at RPI under supervision of Prof. George Saridis spent the academic year 1987-88 in Lund. After his graduation from RPI, and after some time at Fanuc Robotics, Detroit, MI, Steve joined ABB Robotics, first some time at the US office and then participating in the main development in Västerås. Steve's responsibilities at ABB Robotics have included many aspects of control, dynamic models, and software engineering. He is now a chief engineer at ABB Robotics in Gothenburg where he works with so called virtual robots/controllers, digital factories, and the Robot Studio product.
 

In 1992, Ola Dahl presented his thesis on constrained robot control with interesting ideas on trajectory generation. In the same year, Lars Nielsen left our group to take up a position as Professor  in vehicle systems at Linköping University. Soon after, eager to get industrial experience and to try other fields of control, Ola left the department to work as an control systems consultant. In the late '80s,  Gunnar Sparr continued research in computer vision that had started and, in due course, he formed his own, very successful, group on image processing at Dept. Mathematics. This group has maintained an active interest in robotic vision, and recent approaches include an increased amount of collaboration between our departments.

Even though the Irb-6 robot was extensively used in teaching and research, it was rather limited in terms of working range, degrees of freedom, and dynamic properties. In 1990, Klas therefore started NUTEK projects and used his ABB contacts to get a more modern robot, which first was rented and then bought. It was an Irb-2000 with an S3 control system. Earlier at ABB, Klas played a leading role in the early design of the S3 controller, so also in this case he knew  the internal interfaces, and started a major effort with  Rolf Braun in reconfiguring also the new, bigger robot to permit advanced control experiments.

The robotics laboratory now included two robots placed adjacently to one another. Proposed by Lars, the two robots where named Stor-Klas and Lill-Klas (Big Klas and Little Klas after a Scandinavian comics strip), to honor the one that spent, perhaps too many, weeks on the reconfiguration. At this time, however, there was a decreasing interest in robotics within the department. On a longer term, there was even the risk of the laboratory being closed down. But with a genuine interest in robotics, Rolf Johansson and Klas Nilsson decided to continue and to expand. New projects were defined and further efforts in widening the scope of the research was made, as described in the sequel.

The Integrated Robotics Laboratory at Lund (1992--)

The robotics group of  Dept. Mechanical Engineering, headed by Prof  Gunnar Bolmsjö and dating back to 1987, originally had a orientation towards manufacturing gradually developed towards a strong interest in software tools for application programming-in particular, off-line programming and simulation enviroments. At Dept. Industrial Electrical Engineering and Automation, there were activities in electrical drives and sensor technology.  At that department,  Gunnar Lindstedt was working with ultrasonic sensors for object recognition in robotic work cells, and he also helped with programmable hardware for the Irb-2000 interfaces. Over the years, the collaboration with Gunnar and Prof. Gustaf Olsson, previously with  Dept Auomatic Control, has been very rewarding.

In 1993,  coordination of the groups was  stimulated by funding from the Nutek program "Mobile Autonomous Systems"  and Rolf Johansson  was appointed as a coordinator. Anders Robertsson joined the group as a graduate student with a strong background in nonlinear control. Magnus Olsson and Krister Brink and, later, Per Cederberg participated as graduate students at Dept. Mechanical Engineering. Efforts to merge ideas on feedback control and off-line programming were started.  Jan Peter Meeuwse, Bart Hendricks, Olof Laurin made master theses in the Robotics Laboratory under guidance of Klas Nilsson and Rolf Johansson. The research got support from ABB Robotics under the direction of Dr. Torgny Brogårdh. One memorable event in 1996 was the visit of an evaluation committee from Nutek headed by the Program Manager John Graffman and with the veteran robot engineer Joseph Engelberger, retired  manager and designer of Puma robots.

For all  robotic activities, real-time systems have been an important part of the research, and robot control appear to be  among the most realistic case of industrial control implemented within thedepartment. The aim for more flexible and open real-time control systems inspired to new solutions for dynamic linking, as proposed by Anders Blomdell and Klas Nilsson. Another flexible solution, the so called Pålsjö system permitting on-line changes of control software, was developed by Anders Blomdell and Johan Eker. Efforts by Klas Nilsson,  Anders Blomdell, Johan Eker and Olof Laurin provided solutions  for dynamic linking which was a prequisite to on-line change of control Anders Blomdell and Johan Eker developed Pålsjö for programming of tasks of control applications.

Albert-Jan Baerveldt with his PhD from ETH, Zürich, Switzerland  and  Angel Valera with a PhD from UP Valencia,  joined the group as postdoc scientists in 1993 and 1998, respectively.  Norberto Pires with a PhD from Coimbra, Portugal, has been on research visits to the Robotics  Laboratory a couple of different occasions, latest in 1999. The software effort eventually permitted the laboratory to be open -not only to hardware or software components but also to scientists.Robot

In 1996, Klas Nilsson defended his Ph. D. thesis "Open Robot Control Architectures". Based on the observation that robots are distinguished from other types of machinery mainly on the basis of their programmability and ability to be adaptable to different tasks, the  thesis  had a strong emphasis on software engineering. The structure of early control systems, however, limited the applicability of robots, thus leaving many human-unfriendly operations to be performed manually. This thesis took a problem-oriented approach, without enforcing use of formal methods.  Considering industrial demands, such as computing efficiency and simple factory-floor operation, a layered system architecture and technical solutions to accomplish it are proposed. A notion of user views was introduced as the basis for definition of the layers, the layers supporting  programming on levels ranging from implementation of motor control and up to end-user programming. An experimental platform, built around industrially available robots, was developed. Specially developed hardware interfaces and reconfigurations of the original (ABB) system were made to permit control and programming even for lowest-level motion control.

Run-time efficiency within the proposed open and layered system was achieved by a novel concept called actions. Actions were pieces of compiled code that, by use of certain compiling and linking techniques, could be passed as parameters between the layers. The required interplay between application specific programs and built-in motion control could therefore be accomplished. A number of case studies and results from ongoing experimental evaluation indicated that the proposed control system principles were very useful also in industrial contexts.Virtual robot

After Klas Nilsson's dissertation in 1996  and after some months with ABB Robotics, Klas joined the Dept. Computer Science as an assistant professor.  In this capacity, he has maintained an active interest and activity the laboratory. His postgraduate orientation towards real-time systems and safe programming languages--in particular Java technology-is an important asset. The collaboration between control and computer science is active and increasing.

Anders Robertsson Ph. D. Thesis entitled "On Observer-Based Control of Nonlinear Systems-Anders Robertsson" (1999) addressed problems of    observer design and observer-based control for nonlinear systems are addressed, the deterministic continuous-time systems being in focus. A generalization to the observer-backstepping method with the controller designed with respect to estimated states was treated. Moreover, velocity observers with application to mechanical manipulators vehicles were presented. Anders continued his postgraduate career in Robotics Laboratory  with his participation in projects on robotic force control.

In the year 2000, an Irb-2400 robot with an S4C+ controller was kindly made available by courtesy of ABB Robotics. Modifications of the system was this time made together with ABB at their site, maintaining compatibility with the existing system. This development was mainly accomplished by Anders Robertsson (with nicknames Robotsson and Robban) spending many weeks in Västerås.  Following the custom of naming the robot to honor the one who made most of the system engineering, the robot got the nickname "Robban". The new robot was placed in the space belonging to the Dept. Mechanical Engineering, a tangile example of the integrated and multidisciplinary profile of the robotics activities as formed by a relatively small group of people for further successful robotics research in the new millenium.

Research Impact and Industrial Collaboration

ABB Robotics has been remarkably successful and has had for significant time more than half of the world market in heavy robot equipment with a particularly
heavy market penetration in car manufacturing. As for applications, ABB robot systems have proved competitive in arc welding, spot welding, gluing. material handling and machine tending in increased capacity, enhanced quality and flexibility.

Some of the people behind this success have a background from our department. Hans Skoog who later became prominent in engineering and  management at ABB Robotics made his master thesis in 1967 in the area of adaptive control.

The S4C+ is the latest in a long line of ABB robot controllers and it  is designed to match the physical capabilities of the robot range. The S4C+ tries to optimise cycle time and path accuracy  and it tries to be readily integrated into  factory automation plan.  The S4C+  is a compact,  configurable, modular system physically organized with a control cabinet and control pendant are used for all ABB robots.   There are many aspects in common between the work of Klas Nilsson and the current ABB approaches, and the work on open robot control system architectures provides appropriate such  techniques

Currently, cooperation with ABB continues with Klas and Anders frequently visiting for on-site system R&D. We are  happy to stay in touch with the friendly atmosphere at ABB Robotics, and in particular, we are very grateful for the way Dr. Torgny Brogårdh promotes research and long-term development issues within ABB.

Acknowledgement of Project Support

Among important sources of financial support

The   Nutek projects have long been very helpful to our Robotics Laboratory in their support of research with  intellectual quality and industrial application.

References

Doctoral Dissertations

Ola Dahl, Path Constrained Robot Control. PhD thesis TFRT 1038 April 1992. Department of Automatic Control, Lund Institute of Technology, Lund, Sweden.

Lars Nielsen, Simplifications in Visual Servoing, PhD thesis TFRT 1027, Sept 1985, Department of Automatic Control, Lund Institute of Technology, Lund, Sweden.

Klas Nilsson, Industrial Robot Programming, May 29, 1996, Ph.D. Thesis TFRT 1046

Anders Robertsson, On Observer-Based Control of Nonlinear Systems, December 16, 1999, PhD Thesis TFRT 1056,  Robotics Lab, Dept. Automatic Control, Lund University
 

Selected Master Theses

Bart Hendriks: ``Implementation of industrial robot control.'' Master thesis ISRN LUTFD2/TFRT--5555--SE, Department of Automatic Control, Lund Institute of Technology, Lund, Sweden, March 1996.

Olof Laurin: ``Öppna regulatorer för inbyggda system,'' (Open controllers for embedded systems). Master thesis ISRN LUTFD2/TFRT--5528--SE, Department of Automatic Control, Lund Institute of Technology, Lund, Sweden, April 1995.

Jan Peter Meeuwse: ``Algorithms and tools for control of flexible servo systems.'' Master thesis ISRN LUTFD2/TFRT--5531--SE, Department of Automatic Control, Lund Institute of Technology, Lund, Sweden, July 1995.

H. Skoog. Analysis of the Adaptive Control System of Margolis Leondes. Report TFRT-5021, Oct 1967.

E. Mårtensson. ``Active damping of flexible modes in a robot   arm.'' Master's thesis 5359, Dept. Automatic Control, Lund Inst. Technology, 1986. (Awarded ``ASEA MASTER THESIS 1987'')

K. Nilsson. Analysis and Synthesis of the  Dynamics of an Industrial Robot. Report TFRT-5292, March 1983.

Selected Publications

K. Nilsson and R. Johansson. Integrated architecture for industrial robot programming and control. J. Robotics and Autonomous Systems, 29:205--226, 1999.

R. Johansson and M. Spong. Quadratic optimization of impedance control. In Proc. 1994 IEEE Int. Conf. Robotics and Automation, pages 616--621, San Diego, CA, May 8­13 1994.

R. Johansson. Adaptive control of robot manipulator motion.IEEE Transactions on Robotics and Automation,6(4):483--490, 1990.

R. Johansson. Quadratic optimization of motion coordination and control. IEEE Trans. Automatic Control, 35(11):119--1208, 1990.

© Lund Institute of Technology 2004. Web Master.

Last update: Monday, 05-Apr-2004 10:26:19 CEST