Design Of Coverage And Capacity Self-Optimization Platforms And Performance Optimization In Cellular Wireless Networks

With the continuous emergence of new services and applications,higher challenges are imposed on the coverage and capacity of mobile communication networks.In terms of coverage and capacity issue detection,the radio environment map(REM)plays an important role.However,when the radio measurement data is insufficient,the accuracy of traditional REM construction methods suffers significant degradation.On the other hand,to handle coverage and capacity issues,network optimization is required.Although centralized optimization algorithms can achieve good performance,global network information is needed,which can incur huge overhead.Facing the afore-mentioned problems,this thesis proposes an REM construction algorithm based on Kriging with high precision that makes full use of historical data and distributed base station(BS)radio parameter optimization algorithms that just depends on local information.In addition,an experimental hardware platform is designed and developed,to facilitate the collection of real environmental data.The specific contents of the thesis are as follows.Firstly,to improve REM construction accuracy with insufficient measurements,a two-stage algorithm for REM construction based on historical wireless environmental measurements is proposed.Specifically,given the differentiated propagation properties across different zones,historical measurements are utilized to group concerned zones into multiple clusters,in each of which all the zones share the same large-scale propagation parameters.In this way,more accurate REMs can be built.Moreover,when new measurements come,the shadowing effect para-meters in each cluster are further updated via the expectation maximization algorithm,making REMs adaptive to time-varying changes.By training and evaluation of the proposal with real data collected on a software-defined-radio(SDR)platform,it is shown that the proposal can greatly improve construction accuracy compared with other interpolation-based baselines,meanwhile maintain high accuracy when there are few measurements.Secondly,to avoid the overhead incurred by the acquisition of global network information,a distributed reinforcement learning based BS radio parameter adjustment algorithm is developed.Specifically,by the decomposition of Markov Decision Process,each base station can be regarded as an agent that autonomously controls its own radio frequency parameters based on only local information.By interacting with the environment,both the coverage and capacity are optimized.Further,to accelerate the convergence of distributed reinforcement learning,a central controller is introduced to perform cooperative adjustment of multi-BS's parameters.Via simulation,the effectiveness of the proposals is verified compared to centralized schemes.Finally,an experimental platform is developed based on an open-source software,namely OpenAirInterface.Moreover,by developing softwares for users,BSs and the network manager at the application layer,radio environmental data can be collected,stored and visualized conveniently.After the REM construction algorithm is well trained,it can be integrated with the software for the network manager,which can refresh the REMs in real time by receiving measurements of reference signal received power reported by users.