| With the development of telecommunication technologies,there is a big growth in demand of the public for communication quality and capacity.Due to the characteristics of long communication distance,wide coverage of communication beam and good adaption to terrain,satellite communication,which as the supplement and extension to terrestrial communication,can alleviate the pressure from the diversification of communication services effectively.However,some problems like complex channel environment,large transmission delay and limited transmission resources restrict the performance of satellite communication system.How to propose a high-efficiency and high-quality transmission strategy is one of the research hotspots in satellite communication.To solve these problems,channel estimation,channel forecast,and adaptive coding modulation have been proposed.The main works of this paper are as follows:(1)We propose two channel sensing algorithms for in adaptive coding modulation technology:Channel forecast based on time series algorithm with correct term and channel forecast estimation based on deep convolution neural network algorithm.Firstly,we propose a correct algorithm for time series forecast.The simulation results show that the average forecast error of the time series algorithm with compensation term is 1.326dB,which is 40.67%higher than that of the traditional time series algorithm.We propose that using DCNN to estimate and forecast the channel.The effectiveness of these two algorithms is verified by simulation and the result shows that the time series forecast algorithm with correct term has good performance in channel forecast.The SNR estimation and forecast algorithm based on convolutional neural network(DCNN)has good performance in both SNR estimation and channel forecast under C.Loo channel.The results also show that these two algorithms proposed in this paper can effectively solve the problem of inaccurate SNR estimation caused by transmission delay,and probably improve the performance adaptive coded modulation technology.In conclusion,the estimated performance of DCNN is improved by 38.73%compared with the traditional channel estimation algorithm in C.Loo channel.(2)An adaptive coding and modulation control algorithm based on the Double Deep Q learning network(DDQN)algorithm in reinforcement learning is proposed in this thesis.The channel estimation and forecast algorithm based on Deep Convolutional Neural Network is used as the adaptive coding modulation sensing algorithm,and Double Deep Q-learning Network algorithm in reinforcement learning is used as the adaptive coding modulation control algorithm.A new adaptive coding modulation architecture "Double D" based on DCNN+DDQN is proposed.After simulation,the DDQN algorithm as a control algorithm has improved the spectral efficiency by 18.9%compared with the traditional interval method.The new "Double D" architecture for adaptive coding modulation has averagely improved the spectral efficiency by 32.6%and lowered bit error rate by 3 7.5%.(3)We study the DVB-S2 protocol and DVB-RCS2 protocol especially the physical layer framing,coding,modulation,synchronization and decoding part.We build a satellite communication simulation platform whose forward link is based on DVB-S2 and return link is based on DVB-RCS2.We simulate the original adaptive coding modulation technology in DVB-S2 in the simulation platform.The simulation result pointed out that the original adaptive coding modulation algorithm based on the interval method has strategy selection failure under certain SNR conditions. |
