Cognitive radio networks. Assessment framework of cognitive networking for the end-to-end efficiency

Cognitive radio networks are believed to be self-managed. Do we know how to engineer such systems? And, if such engineering is done, do we know how to assess the properties of a self-managed system? Additionally to performance evaluation and testing to guarantee for such systems consistency, safety, purposefulness, security, and efficiency of their operation engineers require assessment - systematic evaluation [of non-conventional to previously engineered systems] properties such as cognition, learning, aware sensing, and perception. Technology-wise we analyse the problem area of flexible spectrum management and suggest that both [self-] management and assessment problems can be solved within the same framework - situation-aware behavioural composition - the key to our policy approach. We describe a model-driven assessment framework: the driving model in the cognition cycle is the enabler of robust control with agile policies - here the "Subject - Object" policy-based self-management framework is introduced; the framework is interpreted from the control theoretical viewpoint; we then compare assessment with the performance evaluation of cognitive networking, and conclude that the evaluation of the process correctness is the main difference between the two. We derive primary assessment metrics and build some derivatives that, using the specific objectivity framework of G. Rasch allow us to define the assessment of process correctness in the way that does not depend on particular cognition algorithm. We report the on-going research and show some early results, followed by the open issues section.