Summary from the Panel Debate at RTiS2003

Embedded Systems of Strategic Importance for Swedish Society -
where are the needs and how should efforts be directed?

Venue and time: Real-Time in Sweden - the 7th biannual Snart conference, August 18-19, 2003, Västerås, Sweden

Introduction

IEEE defines an embedded computer system as a system that "is part of a larger system and performs some of the requirements of that system; for example, a computer system used in an aircraft or rapid transit system". It follows from this definition that computer based systems hidden inside products like for example TVs, telephones, toys and vehicles qualify as embedded systems. Such systems are strongly characterized by their interactions with the environment and requirements on dependable and real-time operation. In addition, because of their embedding they are typically also resource constrained (e.g. due to limited size, power and cost).

Sweden has a strong industrial tradition in advanced control and communication products, including vehicles, telecommunication and process control. The research community has also had a strong boost over the last 15 years decade; here also Snart and the Artes graduate school have played important roles [1,2]. As an indicator of this research growth, five core teams out of some 30 in total, were invited to the Artist2 Network of Excellence proposal in embedded software and systems for the EC 6th framework programme.

The research situation is however commonly seen as quite difficult at this point of time - there is an unbalance in the funding system. This has been expressed by several persons and organizations, for example as follows: "Professors only do 8% of the total amount of research … and spend too much time on looking for funding" - and "Sweden has recently faced a  40% reduction of public research funding".
(Free translations from a statement in Swedish by the General Secretary of the Swedish Science Council, and a survey published in NyTeknik, 25 juni 2003, respectively).

Apart from a general reduction due to the recession, the current funding system is also seen as short-term and fragmented; there are many funding bodies with different priorities and application procedures and the funding is often not granted for longer than three years. Technical universitites have also had their basic granted funding reduced. One overall consequence is that there is less funding for free and basic research, and that in providing means for long term research and critical-mass - much efforts have to be spent on writing applications, evaluations, reiterated too often.

At the same time, it is generally agreed that the importance of embedded systems is steadily increasing - creating new oppportunities in diverse fields such as helth care, equipment for disabled, hand held devices, toys, automotive safety and entertainment systems. For example, the automotive industry claims that 90% of innovation lies in software & electronics in automotive. And, a quick count of the number of "computers" at home, including embedded ones, will - at least for rich countries - underscore the statement that the market size for embedded systems indeed is some two magnitues of order larger than the one for desktop computers.

All in all, embedded systems constitutes an enabling technology that provides opportunities for new companies/business/jobs - thus being essential for economical  and welfare developments in our society.

In view of this current situation, the task given to the panel was centered around the theme Embedded Systems of Strategic Importance for Swedish Society -
where are the needs and how should efforts be directed?

The panelists were asked to provide suggestions and discuss the following

• Their view on the needs and problems related to embedded systems
• How to promote a “successful” research climate to improve the current situation?
• Opportunities?
• Swedish and international perspectives -  What can we learn from other continents?

• Per Skytt – ABB Corporate Research, Head of Division of Automation Technology
• Karl-Einar Sjödin – Vinnova, The The Swedish Agency for Innovation Systems
• Prof. Kang Shin – Michigan/US
• Prof. Heonshik Shin – Seoul/Korea

The introductory presentation by Martin Törngren gave some general remarks on needs - as seen by different  stakeholders - and often suggested key challenges.

The above introduction and the challenges perspective is nicely complemented by the Artes document "Embedded Systems and the Future of Swedish IT-research", [3]. The summary and conclusions part of [3] are restated here: 

Summary: "A major part of Swedish industry is manufacturing products with embedded computer systems. To stay competitive we need continued efforts  strengthening the competence in designing such systems. In particular, since software is the critical factor dominating the design, we need a focused effort into embedded systems software development. The ARTES research network has established an important basis for such an effort."

Conclusions:

• "Embedded systems is the dominating use of computers, and will increase even further as we are entering the era of pervasive computing with massive amounts of co-operating computers controlling virtually all devices in our environment.
• Software is the key component in embedded systems, providing added value and required behaviour.
• Embedded real-time software and systems are major products of Swedish industry.
• The over-all complexity of embedded systems will grow, as they are introduced in new applications, and as requirements on flexibility, integration, external communication etc. are increasing.
• Research and education into embedded systems is vital for Sweden to stay competitive.
• A national research network with associated research projects has proven efficient in strengthening the area."

Brief summary of opinions expressed by the panelists

Characteristics of Embedded Systems (ES)

• Diversity: specific to target machines, domain knowledge required, from small to large. Different drivers; for consumer products (e.g. PDAs): Volume Driver and Technology Drivers emphasized, for Industrial control (e.g. Robotics): Real time Reliability, Safety, Low to large volumes depending on the applications
• Very long life time
• Dominant (in the post-PC market)
• Many new applications are emerging, e.g., Internet and home appliances, entertainment and other gizmos => they are becoming omnipresent!
• Differences from non-ESs: Constrained by cost, weight, size and resources, large volumes, limited in scope, but needs to be more predictable and resilient
• Advantages of ES: Efficient use of resources, effectiveness, Simple, Fast implementation
• Disadvantages of ES: Slow adaptation to changing world, Incompleteness, Dependence on external world

Strategic views: needs and problems

• Embedded software, plus the need to focus efforts
• Embedded systens - not only enabling technology or value adder but also a part of “growth areas”
• Vinnova:  Long term research and cooperation with industry is supported by Vinnova through their competence centres.
• Sweden can not afford to avoid investments related to export of IT intensive products and systems!
• Suggested focus (ABB):
  -Real time and safety critical systems, These areas are needed for main industry like, Volvo, Saab, ABB and Ericsson. Sweden should not focus on ubiquitous computing with computers everywhere.
  
- Main needs include technological (communication, security, reliability, and more see AIC STP) and Organisational issues, where the latter includes difficulties in small organisations or organisations traditionally hardware oriented to handle software development
- Challenges: Interoperability – Standards – Security, Flexibility and maintanability, Development processses and tools in combination with real time and reliability
• In the US: Has had several large programs including QORUM, SenseIT, PAC/C, MoBIES, NEST, PECES, SEC and ARMS, with almost all institutions participating. Coordination could have been better.
• Need to drive innovation at Universities
• More basic research relevant to real world applications AND Need for a balance; thus need for both applied and basic research
• Research needs: Holistic approaches, e.g., HW, OS, middleware, applications, Layperson-friendly interfaces, Packaging, Networking

Suggestions on what to do

• Set up “embedded technology exchange” to help industry solve the problems; consisting of industry, academia, government, research centers and providing a market place for technology. consulting. recruiting. etc.
• Improved education of embedded engineers plus the need to teach and aid in order to drive innovation at Universitites.
• Focus on platforms: Nation-wide supply of embedded software platform including environments: design, development, execution
• There is a need for an improved efficient interaction between industry, academia and funding bodies:
  - More efforts in transitioning existing results and knowledge to industry. Risk funding, which used to be there, is now very scarce.
  
  Innovation needs support by funding and by expertise.
- Industry telling its ES needs to researchers
- Visible impacts from research such as demonstrators
- Improved marketing of research results
- Human resource development including continued education, and education more closely related to industry needs (short and long term)
- Coordination among industry-academia-government-funding agencies; Transfer of research results and education, Focused R&D programs, Committed  industrial participation

Summary from the discussions

 It was agreed that the needs as seen from different stakeholders vary considerably:

• Large companies - would like to see long term research and can adopt research
• Small and medium size enterprieses (SMEs) have little interest in long term research and may not have the capacity/resources for short term research. This motivates specific efforts to encourage interactions between Universities and SME's.
• In common for companies: Need for continued education
• Funding agencies: each have their own program - where the identified needs vary and with different application procedures! Many agencies emphasize application oriented aspects of ES. This implies that it is in many cases possible to get funding for ES research under the umbrella of application oriented research programs.

• While ES constitutes an enabling technology, it is mainly related to applications.
• ES - as such - is a growth area. Many companies can prosper and develop ES specific products such as platforms and tools.

Related to this interpretation it can be noted that ES is one of the prioritized topics in the European Commission 6th framework programme. This should be opposed with Vinnovas currently defined growth areas: Vinnova has defined 18 growth areas of which none is entitled embedded systems. The software products growth area is one that comes fairly close, however many other rely on ES in one form or the other. It is recognized that ES is an enabling technology and Vinnova could, according to Karl-Einar Sjödin, consider to fund ES as such enabling "platforms". As a remark; during the 90's Vinnova had programs directly devoted ES.

Another point made in the discussions was that Sweden has to focus, and select ES related applications and areas where we want to compete world wide. It was also stated that Sweden has to be competitive in brain - not in cost. It was proposed that industry has to get together to discuss focused efforts - an alliance needs to be formed - Sweden is so small that cooperation is needed. Risk funding, which used to be there, is now very scarce. This situation has to be improved. Sweden has to drive innovation at Universities, and innovation needs support by funding and by expertise.

One relevant question in this regard is the empahsis or balancing between a focus on traditional applications and technology - what we are good at now - vs. a focus on new applications.

Prof. Kang Shin mentioned lessons from the US: Visible impacts and marketing are important in research, and Human resources - education - is also very important.

Concluding remarks and an embedded systems characterization attempt

Embedded systems are found in applications with widely varying requirements and constraints such as

• small to large series, implying very different cost constraints, thus different needs for optimization
• relaxed to very strict requirements, and combinations of different quality requirements, for example with respect to safety, reliability, real-time, flexibility and legislation.
  
• short to long life times
• different environmental conditions in terms of for example radiation, vibrations and humidity
• different applications characteristics resulting in static vs. dynamic loads, slow to fast speed , compute vs. interface intensive tasks, and/or combinations thereof.
  

However - the application domains are even broader - and virtually unbounded as pointed out by proponents of Ubiquitous computing, and further mirrored by terms such as pervasive computing, the disappering computer, informative art and smart-its. Such efforts look for new ways of using embedded systems technology [4, 5]. Apart from pointing out new uses of ES such efforts also identify shortcoming and needs such as improving how ES interact with humans and how they can be integrated/embedded into the environment. The use of ES in autonomous robotic systems is another active and interesting field. Some studies try to employ biologically inspired techniqes in the development of ES. Other researchers working on prostheses connect ES to humans - who by the way, it can be argued, contain the most advanced and complex ES known to us. The continued miniaturization of microelectronics further paves the way for new, yet unknown, applications.

This wide variety of ES in effect means that it is very difficult to develop a general view on ES, and calls for a way to characterize or profile embedded systems. Important dimensions, or scopes, of such a characterization could include

1. Requirements scope: This scope defines the qualities being emphasized including refined requirements, mirroring the context of an embedded system. The requirements scope includes basic functional requirements as well as all other conceivable non-functional requirements such as performance (in time and value), reliabilitity, safety, time to market, cost, physical requirements related to weight, size and power, human interface requirements, electromagnetic compatibility, and legal requirements. The requirements scope is clearly essential since it defines the explicit or implicit applicability of any ES approach.
2. Process (phases, activities, guidelines) scope: Efforts in embedded systems are often devoted to a selected number of phases part of the life cycle of ES. For example, tools might be targetting early modelling and concept evaluation or integration testing. In addition, a process may emphasize different activites such as modelling and analysis, or coding, coding rules and testing. Another way to classify activities can be based on the type of solutions that are under design – structural or behavioural solutions, or the mapping between structural and behavioural solutions. The process scope is relevant since it complements the requirements (why) by defining when a given ES approach should be applied, what needs to be done and to some extent how (guidelines).
3. Implementation technology scope: This scope defines the technology used for implementing an embedded system, i.e. how to implement them. ES currently rely on microelectronics and associated development tools. The technology often reflects the application and requirements scope, and may be devised for a broader or a more targetted applications scope. Special processors developed to tolerate radiation is an example of the latter category. Programming languages and tools such as compilers, debuggers, code generators, and test equipment also belong to the implementation technology scope. This scope also includes the technology for packaging and interfacing the ES such as devices used for sensing and human machine interfacing.
   
4. Theoretical scope: This scope encompasses the scientific dimension including developed theories and concepts, such as Maxwells laws, Newtons laws of motion and theories for analysis of properties such as timing and reliability. The theoretical scope includes theories that can be applied to the other dimensions - e.g. a process theory - or alternatively, be treated only in a theoretical context. This scope thus provides support to the other dimensions and provides the fundaments for developing new technology and new processes.

• ES Applications can be seen as a combination of all the above dimensions. An application usually comes along with a set of explicit requirements, may come along with traditional processes or subsets thereof, will definitely employ implementation technology, and may more or less explicitly be based on available theory.
• Industrial technology and tool approaches for ES: Such efforts will typically address the requirements scope, be focussed on one or more phases and activities (item 2), come along with a certain technological scope, and may to a varying degree (more or less explicitly) exploit available theory (e.g. for timing analysis).
• Educational efforts related to ES: Such efforts typically address item 1 above for motivation and exemplification, and in relation to applications (ES scope, requirements and derived constraints). Education may emphasize the process, implemenation scope and/or the theoretical scope. It is quite common that traditional courses have a narrow focus, for example on a particular theory, without touching upon the other scopes. Educational experiences show that relating to the other scopes or by learning theory while tackling the other scopes (so called problem based learning) improves the long term learning process, see for example [7].
• Research efforts related to ES: Here we can distinguish applied research efforts which will have a well defined relation to items 1, 2 and 3, where the actual focus of the research may be on any of the issues or combinations thereof. Basic research topics typically instead focus on item 4, or possibly on long term technological issues (item 4) with a more or less explicit connection to items 1 and  2.
• Processes for ES: Processes should have a well defined relation to the requirements dimensions, should employ and detail how to use relevant theory, and should define the path (top-down and bottom-up) towards an implementation.
• Types of embedded systems: A system type, say distributed, fault-tolerant, power aware or real-time, typically relates to the requirements scope (e.g. real-time and fault-tolerance) and to the process in terms of the considered activities, for example design of distributed systems with an emphasis on structure and/or behavioral issues. The type may relate to the implementation technology through assumptions or by imposing requirements on the technology.

This characterization could be of relevance in comparing different ES approaches and could be one input in discussing focussing efforts.

With respect to the problems raised and perceived by Swedish academia and industry, Professor Kang Shin contrasted with the statement that Sweden is doing very well! Sweden probably has had good funding conditions compared to other countries considering its size and is also industry intensive in this regard. This may well be true - however the panel can be interpreted as concerns raised with how Sweden can maintain and possibly improve its current status.

The situation is serious - an in-efficient research and innovation system will inevitably lead to less research, less well educated people, fewer new companies and less growth of traditional industries in this area - which is already of vital importance for Sweden! We hope that Sweden will be able to build upon its strong industrial traditions and that the current unbalance in the research funding and innovation climate can be improved. To this end there is a need for efforts on the political level!

In the mean-time, the Snart and Artes organizations will continue to work towards achieving these aims, and to discuss and debate the focusing of ES research and education. Embedded Systems are here to stay and countries who master them will prosper!

As an interesting and important side note, the decreased interest in engineering by young people may in the future have a major impact unless handled properly. Our society has never before been more relying on technology, and this dependency will most likely increase in the future. This implies the important need of stimulating the interests of young people - starting at very young ages!

References