Research On Cognitive Engine And Decision-optimization Strategy

The growing of the people’s demand for wireless communication business lead tothe gradually tension of spectrum resource, moreover, fixed spectrum allocation bringsthe low spectrum efficiency. Cognitive radio, as an intelligent spectrum sharingtechnology, renovating the traditional spectrum management system, is one of theeffective methods to solve the problem above. Cognitive engine, as the core technologyof cognitive radio, can dynamically configure its working parameters using artificialintelligence technology according to the changes of communication environment andusers’ requirement. The research focuses on cognitive engine and its decisionoptimization strategy. Contents are as follows:1) Introduced the definition of cognitive engine based on the cognitive cycle, andproposed a new cognitive engine model and described its workflow though comparingand analyzing the character the existing of cognitive engine structure. Based on theanalysis, the main function module of cognitive engine was designed.2) Researching on the input module of cognitive engine, spectrum sensing module.Considering that energy detection algorithm has the shortage in threshold setting, andcyclostationary detection algorithm incarnates high complexity, a kind of adaptivedouble-threshold energy detection algorithm was designed, in which energy detectionwas used outside of the two thresholds to ensure timeliness, and cyclostationarydetection insides, to ensure the accuracy of detection, besides, the two thresholds wasadjusted adaptively according to the channel states. Detection performance of the newalgorithm was simulated and analyzed.3) Researching on the learning, reasoning, optimization and decision module ofcognitive engine. The realization of cognitive decision engine is the process of find thebest solutions in a huge solution space, and artificial intelligence technologies are theeffective way to solve multiple objective optimization problem. Two improvedcognitive engine based on machine learning algorithm were proposed to realize theoptimization configuration of working parameter in multi-carrier system. One wasbased on adaptive immune genetic algorithm, which introduced the adaptive immunegenetic operators into the Immune genetic algorithm, guaranteeing the populationdiversity, overcoming the shortage of weak climbing ability in the later period ofgenetic algorithm. Simulation was made to verify the performance of module in search efficiency, convergence accuracy and optimization performance. The other was basedon binary ant colony simulated annealing algorithm, which introduced the simulatedannealing (SA) algorithm into the binary ant colony optimization (BACO) algorithm,combining the rapid optimization ability of BACO with probability jumping property ofSA, and effectively avoiding the defect of falling into local optimization result ofBACO. Simulation was made to verify performance in the global search ability andaverage fitness.