In heterogeneous frequency demand cellular networks, the frequency demand varies across the cells involved, in contrast to the homogeneous scenario, where the demand is the same. This paper reports the deployment of hybridized evolutionary algorithm and artificial neural network for optimum frequency allocation in heterogeneous frequency requisition mobile networks to ensure interference control. Frequency sharing and reuse among cells, due to scarcity, are fundamental in a communication network for optimum frequency utilization. Primarily, efficient frequency sharing allows cells of adequate reuse distance the utilization of the same channels, with minimized inter-cell interference. The degree of freedom from interference and efficient frequency allocation mechanism dictate the grade of service, GoS, of a communication network. A real-time frequency allocation is defined by some degree of randomness. Thus, frequency allotment problem is mostly expressed as a constrained optimization formulation. The optimal allocations are achieved at points of minimum cost metric. In this paper, the frequency allotment issue is expressed as a two-objective optimization challenge, using Key Performance Indicators (KPIs) data acquired via Drive Test as input parameters. NSGA-II, an evolutionary algorithm was first deployed on the formulated problem, then in combination with SOM, an artificial neural network technique. The hybrid algorithm was implemented in MATLAB for a heterogeneous frequency demand scenario. The results obtained from the hybrid technique show performance improvements of between 6% and 28% in terms of fitness indices for interference cost function and, between 3% and 65% for demand infringement cost function. The algorithm could be embedded in the operating system of Base Station Controllers for enhanced real-time optimal allocation of network resources.
A Channel Selection Algorithm of Power Line Communication Network Base on Double-layer Cascade Artificial Neural Network
