1. Software and Cognitive Radios

In the communications engineering the software radios have become an established paradigm during the last decade or so [Mitola 1995; 1999]. There is a large body of research especially in Europe and in the USA towards realizing more complex and efficient software radio prototypes. The original term of software radio was introduced by J. Mitola in early 1990’s. The basic idea behind the software radio is that increasing computing power and enhanced software algorithms especially for digital signal processing are leading us to a situation in which most of the wireless communication functionalities will be handled by powerful processors running software instead of highly customized hardware. The ultimate software radio would include only smart antennae, AD/DA –circuits, and everything else would be done through software in powerful processing unit. It has pointed out that “As the software radio makes its transition from research to practice, it becomes increasingly important to establish provable properties of the software radio architecture on which product developers and service providers can base technology insertion decisions. Establishing provable properties requires a mathematical perspective on the software radio architecture” [Mitola 1999]. The work on software radio is also conducted quite often under the theme “reconfigurable radio” or “reconfigurable communication technologies”. It has been, of course, generalized to consider not only the wireless radios, but also networking paradigms. In some sense, active networks or programmable networks share a same idea space with software radios.

A far more recent paradigm is so-called cognitive radio introduced by Mitola and Maguire. The basic idea of cognitive radio is deeply embedded with the context sensitivity and context sensitive inference. The context sensitivity in the case of ubiquitous computing and pervasive computing has been a long time known within the computer science community (just to mention the seminal suggestions by Bush, Engelbert and Weiser are enough). The current challenge with cognitive radios is not only technology itself. In the case of cognitive radios one of the challenges is to foster interdisciplinary research, and to understand more clearly what are the specific research problems and boundary conditions that should be met. As an example, it is clear that machine learning and artificial intelligence methods have their important part to play with cognitive radios. Cognitive radio technology is also quite often much harder to achieve than in the case of typical context sensitive applications; stringent real-time requirements and need to co-operate between different protocol stacks make it a complex task, even to model.

2. Cognitive Networks

As the cognitive radio is an extension of software radio (and active networking could be seen as networking counter part of programmable radio), we will speculate with a term cognitive networks . The cognitive network as an idea has been introduced, e.g. by D. Clark [2003] from MIT and Craig Partridge [2003], both well known initial major contributors towards the development of Internet. Clark has described knowledge plane paradigm for Cognitive Networking. Some European researchers have also tried to map challenges and possibilities that could be encountered with wireless and mobile cognitive networks. There has been also some activity emerging from machine learning community in the U.S.A. The cognitive networks have a real change to become a new architectural paradigm change. The current activity, and admittedly scarce, first initial results show that cognitive network might become a very strong research line in the future. It shares in part same challenges as cognitive radios also in the research organization sense. Definitely a lot of initial work is required to nail down the clear research issues and challenges in order to avoid unnecessary “hyped” work.

All of these focus areas are emerging fields (software radios are established, but not really in markets, cognitive radios are rising field, and in the case of cognitive networks we really do not know clearly how to define the boundaries of the research field). Because of these, the seminar could be seen as a slightly unconventional research retreat and workshop. It is not organized “just” to describe and talk about recent results. It is more like (beginning of) exploration – or meeting of explorers to agree where we should go. We aim at in the seminar to

• Define more clearly what should be meant with cognitive radios and networks research.
• To understand what are the research challenges, and what should be done to meet them. (Output should be, e.g., a list on “What are the problems we would like to present to experts in different research fields?” We need to communicate many challenges to other fields from the pure communications sciences and engineering domain. But first, we have to make clear what those challenges are.)
• More specifically; what are the scalability and complexity boundaries for such systems (in the system engineering and network engineering sense), including the realistic possibility for embedded solutions. Many solutions exist already, but in practice they do not scale for communications systems. The more specific questions we need to consider are
• scalability
• architecture and topology constraints
• implementability
• specific research problems esp. cross-layer protocol optimization etc.
• machine learning, optimization and AI related issues
• roadmap for progress, and hardware boundaries.

If we are able to provide some answers and definitions, we feel that we have succeeded. Due to this we are reserving a beginning of the time for a small number of speakers (including a keynote from Chuck Thacker). All the invited members are also welcomed to suggest a contributing paper or short talk to this Dagstuhl seminar.

The main format for our 3,5 day meeting is to have productive workshop and brainstorming. The idea is that collectively we could produce a sort of “consensus report” on the future of cognitive networking, and roadmap towards it. We will have a small number of bright graduate students to make notes during the brainstorming sessions, if so required. Also as a sub-goal would be to discuss that should we have joint research projects to do actual work? The very nice environment and informal atmosphere of Dagstuhl should help on this, and we have reserved plenty of time for informal discussions. A good number of people from industry are also participating in. The informal networking between different academics and some industry people is planned to foster possible future research projects. Results of our brainstorming and all contributed short papers and talks will be collected to seminar report (proceedings) that will be distributed to all participants.