Research On Adaptive Modulation And Coding Technology Based On Reinforcement Learning

In order to cope with the explosive growth of communication service requirements in the future,it is urgent to increase the transmission rate of wireless networks.Link adaptive technology with adaptive modulation and coding(AMC)technology as the core has become one of the most promising directions for improving the efficiency of wireless communication links.AMC is a good solution to match the communication transmission rate with the time-varying fading characteristics of the wireless channel.However,the traditional AMC technology faces the difficulty of configuring parameters such as the target block error rate(BLER)and the channel quality indication(CQI)step size,and has poor flexibility.In recent years,deep reinforcement learning(DRL)has been widely used in wireless communication networks,especially in dynamic decision-making scenarios.In addition,DRL is not affected by configuration parameters such as target BLER and CQI step size,which provides a new method for the effective implementation of AMC technology.In cognitive heterogeneous networks(Het Net)application scenarios,when the spectrum sensing of the secondary user(SU)is inaccurate,the data transmission of the primary user(PU)will be interfered by the SU.Accordingly,thesis proposes a DRL-based resource block(RB)allocation and modulation and coding scheme(MCS)selection algorithm,which solves the mismatch problem between pilot measurement information and RB allocation and MCS selection caused by SU interference.The main research contents are summarized as follows:Firstly,thesis aiming at PU real-time communication requirements and RB allocation fairness,studies the problem of maximizing the total transmission rate of the system under RB one-to-one allocation of PUs,and proposes a DRL-based RB allocation and MCS selection algorithm.Base station(BS)can make more accurate decisions on RB allocation and MCS selection by using the SU interference pattern information learned by the proposed algorithm.The experimental simulation results show that the proposed algorithm greatly improves the total transmission rate of the system compared with the baseline algorithm in quasi-static interference and dynamic interference application scenarios.Secondly,for the problem of high control signaling overhead,thesis studies the performance of the proposed algorithm for RB allocation and MCS selection under different pilot measurement periods.Experimental simulation results show that the proposed algorithm can still outperform the baseline algorithms even with outdated channel information.Finally,for the RB utilization and the different quality of service(Qo S)requirements of different PUs,thesis proposes a multi-agent DRL-based RB allocation and MCS selection algorithm.The algorithm introduces the concept of PU service priority for the agent to guide the agent to allocate more RBs to PUs with high Qo S requirements.The experimental simulation results show that the proposed algorithm can not only meet the different Qo S requirements of different PUs,but also take into account the overall transmission rate performance of the system.