October 15, 2003
Evaluation of the Competence Centre for
Information Systems for Industrial Control and Supervision, ISIS
at
Linköping University
1. Preface, Methodology, and Acknowledgement
On Tuesday, 14 October two of us, the scientific experts of the evaluation team, Michel Gevers and Keith Glover, were briefed by the technical staff of the Centre for Information Systems for Industrial Control and Supervision, ISIS, on their eleven ongoing projects, which are subdivided into their five established activity areas:
Data bases for control, modelling and simulation
Diagnosis, supervision and safety
Techniques for developing integrated control and information systems
Methods for synthesis of control and supervision functions
Signal processing in integrated control and supervision systems
First, the Director, Prof. Lennart Ljung, gave us an overview of the six university teams and the ten companies involved in ISIS, the strategic goals of the Centre and how they were redefined after the mid-term evaluation, and the strategy for project selection. He then gave us a very brief presentation of the eleven presently active projects, indicating when they were started and who the partners are in each project.
The rest of the day consisted of a well-organized “bazaar”, where we were given the opportunity to visit a stall for each project. For each project, we were typically given a brief presentation first, followed by questions, discussions and laboratory or on-the-road demonstrations when appropriate. A number of the industrial partners participated in these presentations and demonstrations. At each of these poster presentations, we had ample opportunity to discuss with the ISIS team members active in this project, including a number of people from the companies involved. We feel that, at the end of the day, we had a very clear view about the different projects, their goals, their state of development and/or the industrial implementations when such have already taken place.
On Wednesday, 15 October, the entire evaluation team, made up of the scientific experts and the two Competence Centre experts, John S. Baras and Per Stenius, had a four-hour meeting with the Director, the Board, and the Reference Group of ISIS. The meeting was briefly addressed by the President of Linköping University, Professor Mille Millnert, who informed us about the decision of the University to create four new chairs in areas that are of direct relevance to ISIS.
We heard presentations about the strategic views of the Centre and its accomplishments by the ISIS Board Chairman Torgny Brogårdh, about the academic perspective of ISIS by the director Lennart Ljung, and about the industrial perspectives by representatives of Ericsson, Mecel, SAAB Aerospace and ABB Robotics, who gave us their views on the importance of ISIS for their respective companies. Prof. Lennart Ljung concluded the presentations with the response of ISIS to the three questions addressed by the generalist evaluators to all Competence Centres. The second day ended with a general discussion between the team of evaluators and the ISIS Board members.
We would like to thank the whole ISIS team for the superb efforts they made in preparing these two days of briefings. The atmosphere that prevails within ISIS allowed for very open and frank expressions of opinions during the discussions. In particular we thank the Director, Prof. Lennart Ljung, for masterminding a very productive use of our time, and the Secretary, Ulla Salaneck, for her superb handling of the administrative details and logistics. We wish to also thank Torbjörn Fängström and Staffan Hjorth of VINNOVA for their thoughtful arrangement and co-ordination that facilitated our evaluations.
2. Technical and Scientific Outcomes
Research programme
The ISIS Competence Centre brings together three groups from the Electrical Engineering and Computer Science Departments with eight ‘flag-ship’ Swedish companies and two SME’s. There is a clearly defined and well-implemented management structure with the ISIS Board (principally from industry), the ISIS Reference Group (principally from the University and meeting monthly) and an International Advisory Board. There is an evolving strategic plan which includes the important constraints on projects that they should both be supported by at least one industrial partner and have the potential to produce scientific research of PhD level. ISIS has been particularly successful at supporting first class projects that meet both of their criteria with both sides well aware of the other’s objectives.
In the application of the above strategy for project selection a new project in Resource Management in Wireless Communication Systems has been added in the 3rd stage in collaboration with Ericsson AB who joined ISIS again in 2000. Also the project entitled Design environment for real-time embedded systems in control related applications will continue under alternative support due to the lack of an industrial supporter amongst the ISIS members in spite of the scientific quality being well recognized.
In the next subsection we first give some outstanding examples of university/industry collaboration.
The mid-term review report of September 2000 had already highlighted the remarkable success of the ISIS management in producing effective collaborations between university and industry in a number of joint projects, leading to the production of both scientific results and PhD theses on the one hand, and to the development of cutting-edge technology on the other. Three areas of successful technology transfer had been pinpointed: navigation and sensor fusion applied to terrain navigation; fault detection, diagnosis and signal processing applied to combustion systems; and iterative learning control (ILC) applied to robot control. These three areas have been pursued and developed very substantially during the third stage: they have either led to actual industrial implementation, sometimes with remarkable commercial success; or the technologies developed in the first and second stage with a particular application in mind have been expanded successfully to new application areas.
In addition to the important new developments in projects whose seeds had already been laid in 2000, the third stage has also seen the emergence of a new and exciting project in resource management in wireless communication systems, with the new partner Ericsson. The interest and investment of Ericsson in this project is very obvious, with much of the present work focusing on network flow control and on control and estimation of uplink load in WCDMA. The mid-term report had encouraged an investment of ISIS in these topics; clearly, this recommendation has been well taken.
We have, in this part of the report, focused on the technical achievements, the implementation aspects, and the technological and commercial benefits to the industrial partners (and hence to Swedish industry) of the ISIS projects.
The Centre has continued to be very productive in publishing its work, and conservative (or honest, we should say) in its reporting by only mentioning work where an ISIS student or post-doc are (co-)authors. The number of concurrent PhD/licentiate students supported is approximately ten, and in the 3rd stage three PhD theses and nine Licentiate theses were completed. This is roughly what would be expected, but the balance since the mid-term evaluation has moved away from PhD theses. On the other hand the number of journal papers has nearly doubled since the last evaluation with a number of research projects coming to fruition, and generally the top ranking journals have been secured. The number of conference papers is large (over 80) and interestingly the number of Master’s theses is also large (57), with many sponsored, for example, by Saab in sensor fusion and by Ericsson on wireless networks. The Master’s theses have clearly formed an integral part of the ISIS research programme in these two areas. Finally, eight patents have been filed, again demonstrating the success of several of the projects.
As a result of the expertise developed within their main areas of competence, ISIS is now recognized worldwide as a leading research group in sensor fusion/particle filtering, in fault detection, diagnosis and isolation, and in Iterative Learning Control applied to robot control. This recognition goes both to the theoretical contributions of the ISIS groups and to the well-publicized applications that have been developed as a result of the university/industry collaborations made possible by the ISIS framework.
Besides the projects, enumerated above, that have led to significant industrial applications during the third stage, a number of other projects have been either continued or initiated during this stage. Some of them are obviously of high quality, even though for some reason or another they may not (yet) have led to major technology transfer to industry.
The investment of ISIS in the control of automobile engines has led them to initiate a project on real-time databases for engine control. This research addresses the problems of data timeliness and transaction scheduling in order to ensure valid calculations with minimal computing requirements. This is a practically important problem, and the project has so far defined a suitable abstraction and identified the strengths and weaknesses of available techniques from computer science.
Fault isolation in object oriented control systems is a project carried on in collaboration with ABB Robotics. It is essentially application-driven and we were informed by ABB Robotics that joint work within ISIS had resulted in their controllers being implemented using object oriented descriptions supported by more formal methods from software engineering.
The goal of the more recent work is to develop a user-friendly diagnosis system that quickly can tell why an industrial robot has made an unplanned stop. This project addresses the problem of informative error reporting in object-oriented control systems. The contribution of the ISIS team has been to produce an automatic abstraction scheme where the system dynamics that play no role in the error propagation are discarded prior to state space exploration, i.e. at a local level. Thus, the fault isolation can be investigated without having to compute the global state space. The theoretical underpinnings of the methods used and how the complexity would scale with problem size were not entirely clear to us, but this might well be due to our lack of expertise in this area. ISIS has also developed some generic tools for fault detection and diagnosis in process control applications, in collaboration with ABB Automation and ABB Corporate Research. One approach is to use dynamic PCA, which is inspired by subspace projection methods. Another approach has been to use structural analysis based on consistency relations obtained from differential-algebraic relations.
Some very exciting work is taking place on robust formulations of Model Predictive Control (MPC) as a step towards the analysis and design of nonlinear MPC. The innovative aspect of this research effort has been to convert the robust formulations of MPC to a framework of minimax performance measures, which lend themselves to linear matrix inequalities and semi definite programming. The goal pursued by the ISIS team is to obtain stability results that do not rely on the commonly used terminal constraints, which are somewhat unrealistic from a practical point of view. We believe that this research topic is particularly promising and potentially useful, given that (linear) MPC is to-date the most successful model-based control design technique used in industrial applications. Although this is not at this point a joint project with an industrial partner, we were told that there is keen interest from ABB Corporate. It is also worth pointing out that the project has led to the development of user-friendly software for Semi-Definite Programming called YALMIP, that is the object of a large number of downloads.
The ISIS PhD students benefit from the creation of the graduate school called ECSEL (Excellence Centre for Computer Science and Systems Engineering in Linköping), funded by the Foundation for Strategic Research (SSF), that aims at providing a broader education across the fields of Computer Science and Systems Engineering. Most ISIS graduate students are enrolled in ECSEL. Since the founders of ISIS are also among the initiators of ECSEL, the goals and competence areas of ECSEL are very consistent with those of ISIS. Several new courses have been created within this framework.
One of the outcomes of ISIS on education has been a massive production of Master’s theses based on collaborative projects between university teams and industrial partners of ISIS. In some projects (sensor fusion and network management, in particular), these Master’s theses have been a major tool for the industry/university collaborations and transfer of knowledge in both directions.
Finally, as a result of the ISIS environment and its effect on student training, the Competence Centre has continued to be an effective platform for industrial recruitment of MSc’s and PhD students. There is now a critical mass of former LiU students who are employed by the industrial partners. One is beginning to observe the long-term beneficial effects of these recruitments for the companies in two ways: the engineers in these companies and their university counterparts now speak a common language, which makes communication much more effective; the relatively large number of young recruits from ISIS bring to these companies an up-to-date know how that is transferred to industry much faster than when such critical mass is not present. We have observed several examples of these beneficial aspects with Ericsson, Nira Dynamics and ABB Robotics.
The increased numbers of journal papers and patents in the third stage demonstrate the effectiveness of the long-term relations between the university teams and the industrial partners within ISIS. They are also a testimony to the balance that the ISIS management aims at keeping between basic scientific research and industrial applications.
In particular the work on diagnosis in vehicle systems; control in robotic systems, and particle filtering in navigation systems have produced academic advances that have had substantial impact on the industrial partners with successful commercial exploitation. This is a remarkably fast transfer from theoretical development to industrial implementation.
We still agree with the Centre’s objective of bringing techniques from systems theory and computer science into closer contact and believe there to be significant opportunities in this avenue. So far the demonstrable success of ISIS in this area has been more modest but the challenges are perhaps greater and we encourage continued endeavour in this direction. In line with this recommendation we feel that ISIS might consider a closer integration of the teams for projects 1 and 5 under the ISIS umbrella. This could assist in creating the required critical mass of researchers aware of the industrial problems in this important but difficult area.
3. Industrial Benefits and Impact on Industrial Partners
Industrial Interaction and Involvement in Centre Activities
ISIS continued its development and improvement as a platform for productive and close collaboration between academic and industry scientists and engineers. There are currently ten industry partners involved in the ISIS programme. They represent a properly balanced mixture of large, medium and small companies from automotive, aerospace, robotics, communications and software systems suppliers and users. The cash support from the companies is adequate but could be increased. The leadership and management of the Centre have succeeded in creating an environment where basic research is pursued that addresses problems of critical significance to industry.
All industry participants were highly appreciative of the benefits derived from their participation in ISIS. These included fundamental enhancements in IT knowledge base and its industrial applications, access to industry partners, access to results and discoveries, productive collaborative research leading to new high value products, access to well educated and trained personnel.
There was ample evidence that personnel mobility between industry and academia was flexible and enhanced by the ISIS environment. All Board members and other industry representatives stressed the value represented in the synergistic working platform they have established within ISIS. All emphasized that it is critical not to lose this unique competence; the current technical theme of ISIS is and will remain critical for the next fifteen years for Swedish industry.
The Board and Centre leadership are to be commended for the persistent efforts and procedures employed in order to establish a commonality of theme and vision. The efforts include workshops to fine-tune the ISIS vision and mission, technical contact groups and the overall ISIS Reference Group. It is a very strong testimony of industry support for ISIS that despite the economic downturn no company has left ISIS.
An important mechanism that has facilitated and enhanced industry-university collaboration and the training of competent personnel for industry has been the strengthening of the MSc Thesis Programme. ISIS has used Master Theses in an innovative strategic fashion. MSc theses work has been directed towards industry relevant projects and thus has provided an important first step for PhD students as well as for students graduating with the MSc degree as their highest degree. From the industrial side, the numerous and high quality MSc theses (57 theses in related areas) have made important contributions to the various industry projects. In addition they have provided many opportunities for students to work on industry problems and interact with industry personnel. This often led to hiring of students by ISIS partners (for example, 15 MSc graduates have been hired by SAAB).
An impressive set of actual commercializations and product development efforts by industry using discoveries from the ISIS programme were presented. It was evident that these successful productizations have resulted in cross-pollination between different companies and have generated new potential applications of ISIS fundamental discoveries towards new products.
The ISIS Board has held discussions and strategic planning sessions regarding the future of ISIS beyond year ten. We highly recommend that these activities need to be intensified and various alternatives need to be carefully examined and evaluated.
Implementation of Results: Technology Transfer, Commercialization, Success Stories.
The research programme of ISIS is focused on information systems and technologies for industrial control and supervision and is organized in five research areas. A strategic plan guides the selection, initiation, and review of projects. Each project must have at least one industry sponsor before it is funded through ISIS.
There have been several impressive technology transfers and commercialization of products utilizing technology and research results from ISIS. In addition functionalities have been added to existing company products and capabilities. To be specific, let us mention the following developments of stage 3.
The collaboration with SAAB Aircraft and SAAB Bofors Dynamics had led to the development of a successful terrain navigation system. This work was combined with more recent work on particle filters and led to the development of a new integrated system for navigation and landing, which was successfully test-flown on the GRIPEN aircraft and will be delivered in the near future to the Swedish Air Force, we were told. This is one of the applications termed “success stories” in the ISIS report.
With this work, the ISIS researchers got a solid experience with the integration of particle filters and sensor fusion techniques, which led to other applications, such as map-based tools for car navigation, or estimation of tire pressure, both applications being commercialized. Thus, NIRA Dynamics offers a TPI (tire pressure indicator), which is based on multilevel sensor fusion. The sensor signals are fused into different model-based filters, and the outputs of these filters are in turn fused to a very robust and high performance tire pressure estimator. This is a cheap software solution with the same performance as much more expensive hardware solutions. It addresses expanding markets; c.f. resulting from US requirements to apply as of November 1, 2003.
The work on fault detection, diagnosis and signal processing in automobile combustion engines, in collaboration with SAAB Automobile and MECEL, had initially focused on on-board diagnosis and engine control algorithms. This work has continued, and in addition the team has moved in the direction of performance improvement for the combustion engine itself based on the estimation of peak pressure position from ion sensing, with the aim of obtaining better fuel efficiency. We were given the opportunity to test-drive cars on which some of the algorithms developed by the ISIS team have been implemented.
The project on supervision and control of industrial robots, in collaboration with ABB Robotics, is the other “success story” claimed by the ISIS management in their report for this evaluation. We have been very impressed by this project, in that it does indeed have all the ingredients of success story. From the start, there has been a very close collaboration between ABB Robotics and the Division of Automatic Control. The initial work on Iterative Learning Control (ILC) has given ABB Robotics a decisive competitive edge, that has allowed it to conquer an important market in the US automobile industry for laser cutting robots in assembly plants, with sales of more than 700 robots using this new ILC technology developed by ISIS. Only one year has passed between successful laboratory implementation and full-scale production; this is exceptionally fast.
ISIS research results and methodologies on fault isolation have resulted in a tool (Dr. Robot) for automatic isolation of the root control problem from a list of error reports generated by the object oriented ABB Robot Controller software. ABB plans to introduce this tool in 2004 in the ABB customer support toolbox.
Finally, the more recent collaboration with Ericsson is focused in radio networks resource management, which is an area where several productization opportunities exist with massive markets. The current emphasis of the ISIS work on WCDMA uplink control provides best possibilities to develop products that can become market differentiators. Three patents have been filed on this technology.
Conclusions and Recommendations
ISIS has developed into a unique source of competence within Sweden in the critical areas of: Sensor Fusion Bayesian techniques and Particle Filters; Iterative Learning Control; Computer Science based techniques for diagnosis; Model based diagnosis; Advanced signal processing in the automotive area. This set of competencies is critical to major Swedish industries, as they testified. Its creation would not have been possible without ISIS. Thus continuation of this productive collaborative platform is essential.
We would like to offer the following recommendations:
Formal Methods and Embedded Systems are critical areas of Computer Systems for the ISIS programme. The current faculty are strong but more faculty positions and more industry-university activities in these areas are needed. The establishment of four new IT chairs presents an excellent opportunity to ameliorate this. We highly recommend that ISIS should vigorously pursue such faculty additions and activity to increase.
A much better balance should be accomplished between projects centred on electrical engineering systems and projects centred on computer sciences. The Board has acknowledged this weakness. A solution will require inclusion of industry partners that are stronger or more appreciative of computer science centred projects.
The Board and Centre leadership are urged to undertake a serious strategic planning exercise regarding the future of ISIS (technical, industry partners including foreign companies, financial) beyond the ten-year mark and to test possible implementation soon.
4. Present Standing of the Centre
The Sensor fusion/Particle filter research group at ISIS is well recognized as one of the internationally leading groups in this area. The work of ISIS in fault detection and isolation (FDI) is also well recognized, and the results of several of the projects within ISIS have achieved international recognition. Thus, the research on information systems for control and supervision at LiU as a whole is of very high standard and is an established participant in the international research network in this area.
As a result of their excellent quality the research groups at ISIS are also demonstrably attractive as a partners in international research projects. For example, they have been invited to participate as partners in five projects within the 6th European framework programmes. ISIS as a research organization has also attracted international interest as a model for industry/university cooperation.
The work at ISIS on sensor fusion and particle filters has created a competence that is unique to Sweden. Much of the work in Sweden on ILC (both academic and industrial) also can be traced back to work at ISIS. Many results of ISIS have been successfully implemented to the benefit of the member companies. ISIS has also contributed strongly to the development of education in IT at LiU. Thus, the Centre has developed a core competence in information systems for control and supervision that has made an impact and should be considered a valuable asset in the development of IT in Sweden.
The Rector of Linköping university (LiU) stated that IT is defined as one of the core areas of LiU in which the university wishes to have a leading position in Sweden. ISIS is a valuable asset as a part of this strategy. The industry/university networks and different ways of collaboration between university, industry and community developed by ISIS and other Competence Centres are of great value to LiU who wish to continue and consolidate this activity, provided that it can also be supported by adequate external funding.
ISIS has contributed strongly to the development of both undergraduate and graduate education in Computer Science (CS) and Electrical Engineering (EE) at LiU. The Centre has been instrumental in creating cross-disciplinary contacts and developed a study profile, “Control and Supervision systems” for undergraduate CS and EE students, with participation of about 25 students per year. ISIS has also participated in the renewal of PhD programmes in CS and EE.
The appreciation of the University contributions to industry products and training of personnel, has led the industry partners to strongly support an increase in faculty resources in areas of direct relevance to ISIS. As a result the University has decided to create four new professorships within the scope of these areas, fully financed by the University: Engineering Databases, Sensor Informatics, Communication and Distributed Systems, Distributed Product Design. LiU directly supports ISIS with a cash contribution of 1 MSEK/year.
Within its area of competence, ISIS has developed to become an internationally recognized research group and a valuable asset to IT technology in Sweden. ISIS has contributed substantially to both research and education on information systems, and this is well recognized by LiU.
We recommend that the support given to ISIS by LiU should be continued and consolidated. It is particularly important that the commitment of LiU to this support is sustained so that the preservation of the core competences created by ISIS beyond the present 10-year VINNOVA financing period is ensured.
5. Future Prospects and Strategies
The development of new knowledge at ISIS and the industrial applications resulting from them have shown that the prospects for future research and path-breaking technological applications in the future are bright.
Industries are internationalizing with concomitant access to research groups in information systems for control and supervision in many different countries. The industrial members felt that ISIS would be able to compete successfully with these and that the close contacts and efficient technology transfer, not in the least by employment of ISIS graduates, should continue irrespective of mergers and ownership.
Strategic discussions with regard to both stage 4 and activities beyond stage 4 have been undertaken and the Board and the reference group of ISIS have performed a SWOT analysis. The weaknesses in computer science and database management noted in the previous evaluation of ISIS will to some extent be alleviated by the creation of the four new IT-related chairs. A weakness of ISIS is still its dependence on relatively few core persons, both at ISIS itself and in the supporting industries. The research programme was felt to be very strong and worthy of further development.
Accordingly, researchers at ISIS and the supporting industries agreed vehemently that the development at ISIS has been very successful and that research at ISIS in the future should be focused on development and deepening of the present areas rather than extended into new ones. Based on this strategy, ISIS will investigate different routes for external support, with continued support in the present form from VINNOVA, industry and university as the most desirable alternative.
Technological and scientific prospects in the ISIS research area are bright; ISIS is well recognized nationally and internationally and so far has received strong support from industry and university. The conclusion to continue the form of support along the present lines, therefore, makes sense in terms of the benefits to Swedish industry and the we feel that this could be a viable solution.
Competition for support to Competence Centres and similar cooperative industry-university ventures will be fierce in the future, and development of research and applications in information systems is very rapid. We therefore feel that, in spite of the very high quality of the present activity at ISIS, it may be advisable to consider also alternative solutions.
The third stage of ISIS has seen a significant deepening of the university/industry collaboration. This has led to remarkably successful examples of technology transfer, which have in several cases been a key factor for industrial and commercial success for the participating Swedish companies, as highlighted in Section 3 of this report.
The increased number of journal papers and patents in the third stage also demonstrates the effectiveness of the long-term relations between the university teams and the industrial partners within ISIS. They are also a testimony to the balance that the ISIS management aims at keeping between basic scientific research and industrial applications.
ISIS has developed into a unique source of competence within Sweden in the critical areas of: Sensor fusion Bayesian techniques and particle filters; Iterative learning control; Computer science based techniques for diagnosis; Model-based diagnosis; Advanced signal processing in the automotive area. This set of competences is critical to major Swedish industries, as they testified. Its creation would not have been possible without ISIS. Thus continuation of this productive collaborative platform is essential.
We wish to make the following recommendations:
ISIS has so far suffered from a certain imbalance between projects centred on control systems and projects centred on computer sciences. The Board has acknowledged this weakness, and the efforts to establish a better balance between projects centred on electrical engineering systems and projects centred on computer sciences should be continued. The establishment of four new IT-related chairs is to be commended and should be fully exploited towards this end. A satisfactory solution may require in addition inclusion of industry partners that are more appreciative of computer science centred projects.
Technological and scientific prospects in the ISIS programme are bright. The overall ISIS competence is critical for Swedish industry. Thus continuing support along similar lines and mode provide a reasonable and viable solution. Nevertheless, given the rapidly evolving development of research and applications in information systems, we recommend that the Board, the reference group and the International Advisory Board of ISIS examine in the near future alternative ways to ensure the sustainable development of the education, research and industrial network created by ISIS.