Research On Adaptive Modulation Coding And Power Allocation Algorithm Based On Supervised Learning Method

In complicated MIMO-OFDM systems,performances of the system are determined by the channel state.The traditional feedback adaptive modulation coding problem essentially switches the MCS by establishing a lookup table with feedback information such as channel matrix and SNR information.By choosing a MCS,state effects are mapped to system performances.This type of mapping to establish the channel state to system performance is complex and non-linear.At present,there is no specific theory or method to determine such mappings.However,the learning algorithm is effective on the processing of nonlinear mapping.This kind of mapping relationship can be reflected in the model by establishing a relationship between input and output.Therefore,more and more scholars begin to explore the use of learning methods to deal with adaptive control problems.This is an alternative strategy when there is no specific theory to model such complex nonlinear mappings.Supervised learning is a big category of machine learning algorithms,which can achieve good performance in solving classification problems.The progress of choosing modulation and coding schemes for different channel realizations can be regarded as a typical classification problem.Therefore,an adaptive modulation and coding algorithm based on supervised learning is proposed,and power allocation method based on the proposed modulation and coding algorithm is then discussed to improve the overall data transmission rate performance of the system.The adaptive modulation coding algorithm based on supervised learning is first proposed.This algorithm is without feedback.Two typical learning algorithms,K nearest neighbor algorithm and neural network are implied to solve the classification problem.In order to transform the adaptive modulation and coding problems into classification problems,this paper proposes a channel feature extraction scheme based on channel conditions and transmission power constraints,which is the basis of the whole paper.Then the feature extraction scheme is applied into practice for different modulation and coding schemes to verify whether the feature extraction scheme can accurately classify different modulation and coding schemes.Next,in order to avoid dimensional disasters and to represent different spatial stream accurately,two kinds of dimensionality reduction schemes are proposed for the proposed feature extraction scheme,and the feasibility analysis of the two kinds of dimensionality reduction schemes is carried out.Then,on the basis of feature extraction and dimension reduction,K nearest neighbor algorithm and neural network algorithm are added to the adaptive modulation and coding problem,and KNN-AMC algorithm and ANN-AMC algorithm are proposed.Finally,KNN-AMC algorithm and ANN-AMC algorithm are applied into simulation.The feasibility of KNN-AMC algorithm and ANN-AMC algorithm is verified and their performance is analyzed by analyzing the bit error performance and data transmission rate of the system.In LTE TDD system,channel information is obtained by utilizing the reciprocity of uplink and downlink channels.It can be used in the algorithm model of this paper to save feedback overhead and reduce system working delay,which can make the system adapt to the change of communication environment more quickly.The second problem that this paper focuses on is the power allocation problem based on KNN-AMC algorithm.At first,the traditional Water-Filling algorithm is used to solve the power allocation problem.Through the analysis of data transmission rate,the Water-Filling algorithm can improve the data transmission rate of the system.Considering that the Water-Filling algorithm is a method to adjust the process of bit-loading to reach the optimal value,which means different streams need to be modulated in different modulation shcemes.But the adaptive modulation and coding scheme selection performs the same modulation scheme for all spatial streams in this paper.Therefore,this paper proposes a power priority allocation algorithm to improve data transmission rate.The main idea is to collect the residual power first,and then to distribute the residual power according to the modulation level from high to low.Finally,the priority allocation algorithm is compared with the water-Filling algorithm through simulation.And it can be seen that the priority allocation algorithm achieves better system transmission rate performance through discrete optimization of the Water-Filling algorithm.