Coverage And Capacity Optimization Research Based On Computational Intelligence In Mobile Communication Network

The technologies of coverage and capacity optimization are important to the mobile communication networks.In some scenarios,ranging from ultral dense base stations,three-dimensional coverage to dynamic network enviroment,the coverage and capacity optimization proplems are possibly with nondeterministic polynomial time complexity.Therefore,in order to overcome the aforementioned problems,a series of efficient computational intelligence algorithms are proposed.The main research outcomes and innovations are listed as follows:(1)Aiming at the requirement of coverage optimizationl in ultra dense network,we explore the characteristics of the shortest arc and equivalent permutation of antenna azimuth variables,construct the quotient space of azimuth variables and propose evolutionary algorithm and particle swarm optimization algorithm adapted to the solution space structure of variables.Simulation results show that the proposed evolutionary algorithm and particle swarm optimization algorithm outperform other baseline algorithms with respect to the convergence speed and the final coverage,and have advantages of stability compared with the corresponding canonical algorithms.(2)In view of requirements of ground and three-dimensional coverage,we first get insight of the intrinsic space structure of antenna orientation variables.Then,we propose the quaternion-based particle swarm optimization algorithms,in which the quaternions represent the antenna orientations,and the rotations formed by the multiplication in the quaternion field ensures that the feasible solutions update along the shortest paths.Moreover,the blending quaternion-based particle swarm optimization algorithm based on infinitesimal movements is further introduced to overcome the problems caused by the anticommutative law of quaternions.Last,this paper proposes Nelder-Mead simplex algorithm and genetic algorithm based on quotient space of variables to optimize network coverage.Simulations show that the proposed particle swarm algorithms,Nelder-Mead algorithm and genetic algorithm perform better than the canonical computational intelligence algorithms both in the convergence efficiency and the final optimized coverage.The computational complexity analysis shows that the proposed algorithms can significantly improve the coverage performance with limited additional computational overhead.(3)For the capacity optimization problem of cooperative resource allocation,we consider the cellular network scenario of dynamic time-varying channel state and energy-constrained multi-base stations.First,the resource allocation problem is modeled as the policy optimization problem of deep reinforcement learning.Then,the equivalent transformation relation of optimization problem is analyzed in detail.Last,we propose the data augmentation schemes based on equivalent permutation and devise two different augmented schemes for repaly buffer of deterministic and stochastic strategy gradient algorithms,which effectively improve the diversity of repaly buffer and reduce the cost of interacting with environment.Simulation experiments show that the proposed algorithms not only converge rapidly but also improve the availability of algorithms in real dynamic environments.In this way,the capacity of the system can be maximized while taking into account the fairness,and the cooperative resources allocation problems are effectively settled in the dynamic network environment.In conclusion,this research aims at the coverage and capacity optimization problems in mobile communication network.The solution space structure of decision variables is intentively studied to boost the efficiency of the computational intelligence algorithms,which will provide technical support for the coverage and capacity optimization in the future mobile communication networks.