Dynamic Cross-layer Routing Using Cognitive Spectrum Allocation Techniques

ABSTRACT: The CTI/SU team proposes the Cyber Cross-Layer Optimization Publish-Subscribe (CYCLOPS) framework for design, prototyping and assessment of innovative methods for creating cognitive network architectures and protocols to achieve autonomous network resiliency in contested Radio Frequency spectra. Within this framework, the team will design metrics for estimating the benefits, costs and risks (vulnerabilities) associated with alternative wireless network Courses Of Action (COAs). Within this framework, the team will prototype both conventional filters and cognitive agents to perform multi-objective (benefits, costs, risks) optimization that will dynamically select COAs, including spectrum allocations, network routes, [reliable] [multicast] transport parameters and Delay/Disruption/Disconnection Tolerant Network (DTN) bundle transport options. Within this framework, loosely coupled agents created by different methods (design by human engineers, evolution by genetic programming) and tuned by different methods (learning neural network weights, learning classifier system rules) will interoperate synergistically by exchanging information. The heart of the CYCLOPS framework is a distributed blackboard in the management plane: control/user plane protocol stack layers and other entities, such as battery level monitors and Global Positioning System (GPS) receivers, publish information of potential value; other entities, especially stack layers, subscribe to relevant information and use it to optimize their behavior. BENEFIT: Through the design and development of an automated cross layer information sharing blackboard with a publish/subscribe architecture and an automated, cognitive, optimization decision optimization algorithm, this solution will determine costs, risks, and benefits for the sharing of state data across the layers, and for the execution of any resultant optimization decision. Oracles based upon predictive and cognitive, evolutionary learning algorithms will identify and evaluate the current and near-future node state, integrate individual function, node, AND system policies, costs, risks, and benefits, and then generate decisions which approach long term node and network optimization. In addition to addressing these questions and developing an optimization decision, these efforts will instantiate the concept of executing a decision NOT to share information and of NOT to"optimize". Thus individual agents will execute and report on their own OODA (observe, orient-assess, decide, & act) loop and optimization options, and system agents"will evaluate the net cost/benefit effect of the decisions of each individual agent; identify decisions that yield maximum benefit/minimized cost/managed risk; and will then also eliminate individual optimizations that result in minimized gains but increased, undesirable long-term system costs (e.g. thrashing between channels for minimal additional throughput). Given a selection of any mobile wireless platform based upon the Android 2.0+ OS, CTI and SU will plan on transitioning this technology as a series of licensure products customized for each wireless handheld manufacturer. The initial customer will be the US wireless providers deploying the Android-based OS in their"smart phone"offerings, whom have already been identified during CTI"s previous commercial program (the Personal Status Monitoring Project)