Research On Resource Allocation And Optimization For Large-Scale Internet Of Things Communications

In recent years,with the rapid development of Internet of Things(Io T)applications(such as smart home,smart grid,smart factory,etc.)and wireless cellular communication technology(e.g.,5G and beyond),the number of Io T devices has increased exponentially.It is predicted that by 2030,10 million Io T devices per square kilometer will need to be connected to Internet,among which the number of Machine-Type Communication(MTC)devices will soon exceed that of Human-Type Communication(HTC)devices and become the mainstream.Unlike HTC,MTC usually has the characteristics of massive connection,sporadic transmission,small packet,bursty traffic,limited battery as well as diverse Quality of Service(Qo S)requirements,and the traffic of MTC is mainly concentrated in the uplink.However,the current cellular network is mainly designed for HTC,and it is not ideal to directly integrate MTC into the existing cellular network.Simultaneous access of a large number of MTC devices will cause network congestion and have a great impact on HTC.Therefore,it is necessary to construct a communication system that can satisfy the different characteristics of HTC and MTC at the same time.Grant-free access,ultra-dense networks,and Non-Orthogonal Multiple Access(NOMA)technologies are expected to enable coexistence of HTC and MTC,but realizing performance gains through these technologies also requires tackling some challenges.Therefore,for large-scale Io T communications,this dissertation investigates the device transmission resource selection mechanism from the perspective of devices,and investigates the Small Base Station(SBS)state selection strategy as well as NOMA power level pool design scheme from the perspective of Base Stations(BSs).The main contributions of this dissertation are as follows.1.Due to the characteristics of small packet,bursty traffic or probabilistic traffic,grant-free access and distributed transmission resource selection are more suitable for MTC.However,for a single MTC device,designing distributed transmission resource selection method to mitigate transmission collisions with minimal reliance on historical experience is a significant challenge.To address the collision problem caused by a large number of devices using grant-free access,this dissertation combines grant-free access with code-domain NOMA technology to study the uplink massive MTC scenario with probabilistic traffic,and proposes a distributed Q-learning based transmission resource selection scheme from the device perspective.The idea of MTC device grouping as well as transmission resource grouping,and intermittent learning mode are adopted to accelerate algorithm convergence,and to alleviate the discontinuity of agent learning caused by MTC traffic characteristics.The proposed scheme shows good overall performances compared to the benchmarks.Additionally,the proposed scheme is applicable to scenarios with high device density.2.Although dense deployment of SBSs can provide more access opportunities for devices,compared to the single BS scenario,ultra-dense network introduces resource management at the SBS level.Keeping too many SBSs active can meet the Qo S requirements of HTC and MTC devices,but it will increase the economic and environmental burden.Shutting down too many SBSs can reduce costs,but it may not be enough to meet the requirements of devices.Therefore,how to determine the SBS state(activated or deactivated)in an ultradense network needs to be carefully considered.Aiming at the resource allocation problem in the ultra-dense network where HTC and MTC coexist,this dissertation investigates the SBS state selection strategy from the perspective of BSs.Considering the different characteristics of HTC and MTC,HTC device and MTC device adopt grant-based method and grant-free method to access SBS,respectively.This dissertation proposes an SBS state selection scheme based on multi-agent deep Q-learning,in which each SBS as an agent selects an appropriate state between activated and deactivated through continuous interaction with the environment,so as to achieve the goal of using as few SBSs as possible to serve as many devices as possible.The proposed scheme exhibits superior performance in terms of packet successful transmission probability and average BS load.3.The NOMA power level pool is a key component for power-domain NOMA to bring performance gains,but there are few works on the design of NOMA power level pool and these studies are mainly conducted in single-cell scenarios without inter-cell interference.In ultra-dense networks,inter-cell interference cannot be ignored,and when BS has imperfect Successive Interference Cancellation(SIC)capability,co-channel interference will be more complex.This further increases the complexity of designing the NOMA power level pool and renders the solution used in single-cell scenarios no longer applicable.In this dissertation,we first study the NOMA power level pool design in the presence of inter-cell interference,and propose a non-orthogonal random access scheme assisted by NOMA and tagged preamble technologies,in which the hybrid strategy that integrates centralized power allocation and distributed power control is adopted to determine the transmit power of devices.Then a multi-agent deep reinforcement learning algorithm is presented to optimize the NOMA power level pool for each SBS,and the situation that SBSs have imperfect SIC capability is considered,thereby further empowering HTC and MTC to coexist in ultra-dense networks.The proposed scheme can achieve similar performances as the brute-force search scheme in terms of MTC packet success probability and average delay,but there is still a gap in terms of HTC average rate.In summary,the main contributions of this dissertation are to propose device transmission resource selection scheme,SBS state selection scheme,and NOMA power level pool design scheme for large-scale Io T communications.Through these investigations,a better understanding can be obtained of the series of problems that need to be addressed for the coexistence of HTC and MTC,as well as how to solve these problems through new technologies,thereby providing theoretical basis and technical support for the design,construction and optimization of future cellular communication systems.