Random neural network based power controller for inter-cell interference coordination in LTE-UL

This paper presents an application of cognitive networking paradigm to the problem of inter-cell interference coordination (ICIC) in Long-Term Evolution-Uplink (LTE-UL). We describe state-of-the-art, research challenges involved, and a novel random neural network (RNN) based power controller and interference management framework. The RNN based cognitive engine (CE) learns how the electromagnetic environment is affecting the reliability of communication and can therefore dynamically selects the optimal transmit power for serving users and the users' served by neighbouring cells. The presented CE model has features such as better generalization, fast calculation in run-time, and easy hardware/software implementation. These three features are essential for the development of any real-time cognitive communication system and most of the existing artificial intelligence (AI)/machine learning (ML) based solutions lack them. To achieve the best learning capability, performance evaluation of four learning algorithms (gradient descent (GD), adaptive inertia weight particle swarm optimization (AIW-PSO), differential evolution (DE), and genetic algorithm (GA)) are applied to RNN and analysed. The analysis showed that AIW-PSO outperformed all other algorithms in terms of both learning and prediction accuracy but with slight compromise on computational complexity and training speed as compared to GD, whereas DE and GA had low performance. Therefore, AIW-PSO based RNN framework is the optimal choice for presented system model.