Self-configuration and optimization for cognitive networked devices

Connected devices are an integral part of the ubiquitous and pervasive systems. They are highly dynamic and can grow into complex networks with multiple connectivity options, application requirements and user contexts. Since manual configuration and optimization are tedious for such systems, autonomic management would be beneficial. This motivates the conceptualization of cognitive networked devices that follows a cognitive approach for management. One of the major architectural components considered here is a cognitive management entity supporting self configuration and self optimization with an end-to-end perspective. Here we explore an optimizing controller which follows a cognitive process that can run in real-time efficiently on resource constrained device platforms in the absence of a mathematically tractable model. A methodology is developed based on dependency graph as an ontology modeling relative influences between state variables. Concepts from monotonic influence graphs and signed graphs have been used to formulate an approach and algorithm for a qualitative optimization. Functionality of the approach has been demonstrated with an example.