Research On Optimization Of Resource Allocation And Low Complexity Signal Detection In SM-MIO Vehicular System

With the rapid development of wireless communication technology,new services and businesses have emerged constantly,leading to explosive growth in mobile data traffic and posing significant challenges for the fifth-generation(5G)and future wireless communications.As an essential scenario in 5G and intelligent transportation systems(ITS),vehicular communications have attracted great attention to enhance road safety and traffic flow through high data rates,low latency,and high bandwidth.In vehicular communications,increasing data traffic may cause overcrowding of the available frequency band.Thus,wireless communication systems require emerging communication technologies to achieve the goal of high spectral efficiency(SE),energy efficiency(EE),and reliability.Multiple-input multiple-output(MIMO)technology has emerged as a potential transmission technique in 5G vehicular communications and beyond 5G.However,the use of multiple antennas introduces challenges such as interchannel interference(ICI),inter-antenna synchronization(IAS),and power consumption.Furthermore,the high Doppler propagation in the vehicular environment would aggravate the spatial correlation of transceiver antennas,reduce the achievable multi-antenna gain,and lead to a challenge for large antenna technologies in vehicular communications.To overcome these challenges,spatial modulation(SM)is a highly promising solution for the application of the MIMO system since it achieves excellent balance between SE and EE.SM extends the traditional 2-dimensions digital modulation to 3-dimensions by adding a spatial dimension and utilizes antenna activation(TA)selection for information transmission.However,in the high mobility scenarios,the typical SM technology are highly sensitive to dynamic channel characteristics and serious interference.Thus,the combination of cooperative technology and SM provides an efficient solution for vehicular communications.Furthermore,the typical SM system may not fully exploit the freedom degrees offered by the MIMO system and lack the transmit diversity.To deal with these issues,SM technology is combined with link adaptation(LA)techniques,where the transmission parameters can be adapted to the varying channel conditions.Specifically,power allocation(PA)and transmit antenna subset selection(TASS)techniques are viewed as efficient resource allocation(RA)based LA schemes in the spatial modulation multiple-input multiple-output(SM-MIMO)system.Nevertheless,most of the LA schemes are commonly solved by high complexity.Also,the traditional SM detection algorithms face a severe imbalance between computational complexity and detection performance.Moreover,due to the limited wireless spectrum resources,system capacity and EE improvements are still critical challenges for SM-MIMO vehicular system.Thus,effective resource block assignment(RBA),PA,TASS,and signal detection(SD)techniques are important to improve the performance of the SM-MIMO vehicular system with low complexity.The rapid development of machine learning(ML)and deep learning(DL)technologies has provided new opportunities to investigate effective intelligent transmission techniques,SD algorithms,and RA algorithms in complicated vehicular communication environments.Although high mobility,changing channel conditions,and high quality of service(QoS)requirements,these advanced technologies can improve the performance of vehicular communications significantly while reducing the computational complexity.Based on the above analysis,this thesis mainly focuses on RA in the SM-MIMO vehicular systems with low complexity SD by designing high efficient PA,TASS,and RBA algorithms to adapt the transmission parameters to the channel conditions,and provide effective complexity-performance tradeoff.The main research contents of this thesis are summarized as follows:1)This thesis proposes a low complexity adaptive PA in the SM-MIMO vehicular system to mitigate the impact of MIMO fading channels and enhance the communication performance.We first evaluate the exhaustive search Euclidean distance(ED)based PA(ED-PA)scheme and then propose the error vector magnitude(EVM)based ED-PA algorithm.Specifically,a simplified closed-form PA strategy for the quadrature amplitude modulation(QAM)is obtained in the case of two TAs.In contrast,for large TAs,a low-complexity PA(LC-PA)algorithm is further proposed,where the TAs are grouped with two TAs in each group.Then,the proposed EVM based ED-PA is employed to assign the power within each group.Numerical results show that the proposed EVM based ED-PA and LC-PA algorithms achieve a significant average bit error rate(ABER)performance improvement with lower complexity compared with the conventional PA counterparts in the SM-MIMO vehicular system.Considering the advantages of the proposed PA algorithms in the SM-MIMO vehicular system,they are further employed in the cooperative SM-MIMO(CSM-MIMO)vehicular system to improve the overall system performance.Numerical results demonstrate the advantages of the proposed PA schemes in the CSM-MIMO vehicular system.2)To further improve the vehicular system performance and provide high transmit diversity,low complexity TASS algorithm is proposed in the SM-MIMO vehicular system.This thesis presents a disjoint TASS strategy to reduce the computational complexity of the TASS in the SM-MIMO vehicular system,where the possible TA subsets can be determined by dividing the total TA numbers by the number of the selected TAs.Then,the TA subset that maximizes the minimum squared ED(MSED)at the receiver is selected for the transmission.Also,two low complexity TASS algorithms,TASS algorithm I and TASS algorithm Ⅱ,are proposed for the CSM-MIMO vehicular system.For the TASS algorithm I,the relay selects the best TA subset via maximizing the MSED for the transmission to the destination.The relay then feedbacks the index of the chosen TA subset to the source.This TASS algorithm requires no complete channel state information(CSI)at the destination and substantially reduces the computational complexity of the cooperative system.For the TASS algorithm Ⅱ,the TA subsets at the source and the relay that maximize the MSEDs can be respectively selected,which improves the system performance by considering the CSI conditions of the source-to-relay and relay-todestination links.Finally,closed-form ABER expressions for the SM-MIMO and CSM-MIMO vehicular systems with the TASS algorithms are derived based on the union bounding techniques and order statistics.Numerical results demonstrate the correctness of the derived theoretical ABER.Also,the ABER performance of the proposed TASS in SM-MIMO and CSM-MIMO vehicular systems outperforms the typical counterparts without TASS.Additionally,compared to typical TASS algorithms under different vehicular parameters,the proposed TASS algorithm in this paper significantly improves ABER performance in the svstem.3)The traditional TASS based decision-making optimization algorithms have limitations in balancing computational complexity and system performance.Although the proposed TASS algorithms in Chapter 3 can reduce the system complexity,they may give a suboptimal ABER performance with lower transmit diversity compared with the exhaustive search based TASS(ES-TASS)algorithm.Thus,this thesis further proposes the supervised learning classifier(SLC)based TASS(SLC-TASS)for the SM-MIMO vehicular system to overcome this problem.Specifically,the ES-TASS problem is transformed into a classification problem,and a low dimension multi-output classifier is designed to achieve the low complexity solution of TASS.Then,two SLC-TASS schemes,namely random forest decision(RFD)based TASS(RFDTASS)and deep neural network(DNN)based TASS(DNN-TASS),are proposed to choose the suitable TA subset.Moreover,to overcome the issue of high detection complexity and improve the overall system performance,DNN based SD(DNN-SD)scheme is proposed for the TASS of the SM-MIMO vehicular system.Specifically,two subDNNs are developed to recover the transmitted SM signals.The proposed SLC-TASS and DNN-SD schemes are trained individually under static vehicular motion.Numerical results reveal that the proposed DNNTASS scheme attains better ABER performance than the RFD-TASS and outperforms the traditional TASS schemes with lower complexity in the SM-MIMO system under different vehicular parameters.Furthermore,the proposed DNN-SD scheme obtains a superior detection performance compared with the conventional linear detection methods and provides the same ABER performance as the optimum maximum likelihood detection(MLD)scheme in the presence of correlated noise.Considering the advantages of the proposed DNN schemes in the SM-MIMO vehicular system,the proposed DNN-TASS and DNN-SD schemes are employed in the CSM-MIMO vehicular system.Numerical results reveal the superiority of the DNNTASS and DNN-SD schemes in the CSM-MIMO system under different vehicle parameters.4)Due to the inefficient RA and high QoS requirements in vehicular communications,multiple V2V links sharing the resource blocks(RBs)can lead to high interference and degrade the system capacity and EE.Therefore,this thesis proposes two dynamic RA algorithms,spectral clustering based greedy(SCGR)algorithm and multi-agent deep reinforcement learning(DRL)algorithm,to maximize both the sum capacity of the vehicle-to-infrastructure(V2I)uplinks and the total EE of the vehicle-to-vehicle(V2V)links by assigning the proper power and RB to each V2V link in the SM-MIMO vehicular system.For the SCGR algorithm,the spectral clustering(SC)scheme is first utilized to group the V2V links for the suitable RBs.Then,the optimal power is distributed to each V2V link by the greedy(GR)algorithm.For the DRL algorithm,a decentralized model-free network,improved multi-agent deep Q fully connected neural network(IDQFN),is developed to simultaneously find the best PA and RBA.Numerical results reveal that the proposed SCGR and IDQFN RA schemes outperform the typical RA algorithms significantly,and the IDQFN scheme achieves better EE than the SCGR scheme,while the SCGR algorithm obtains the optimal ABER performance.