Research On Resource Allocation For Low Power Internet Of Things

With the development of wireless sensing and communication technologies,low power Internet-of-Things(Io T)is widely deployed to collect environmental data.Diverse application scenarios and the increasing Io T devices pose challenges to the performance of data transmission in low power Io T.On the one hand,the conflict between the explosive growth of Io T devices and limited wireless resources has become the bottleneck that impacts the performance of low power Io T.Therefore,allcating the limited wireless resources to support reliable data transmission of a large number of Io T devices has been an important research problem.The resources that can be allocated mainly include wireless channels(Allocating different channels for Io T devices to avoid signal conflicts),spreading factors(For Io T systems that use linear spread spectrum modulation,Allocating spreading factors to achieve different signal transmission reliability and energy consumption),time slots(By dividing the time into small time slots,assign different time slots to Io T devices to avoid signal conflicts),and signal transmission power(Io T devices with different transmission power have different signal transmission reliability and energy consumption),etc.On the other hand,due to the diversity of application requirements,low power Io T networks may be deployed in scenarios with different characteristics.For example,in urban areas,dense buildings and other obstacles will shadow the transmission of wireless signals,and the devices are often clustered in buildings and requires transmission reliability.However,in suburban areas,the Io T devices are often sparsely deployed,and due to the high cost of manually replacing the batteries of devices,network lifetime is critical for Io T networks.To this end,this thesis studies the factors affecting the performance of low power Io T systems(such as transmission reliability,network throughput and resource utilization,etc.)under different wireless communication technologies for different scenarios,and studies the corresponding low power Io T resource allocation schemes.Specifically,the main research content of this dissertation are as follows:(1)Lo Ra networks,as Low-power Wide-Area Networks(LPWANs),has been widely deployed to cover wide areas due to its long signal transmission distance and low energy consumption on devices.When the batteries of the devices are exhausted,the cost of manually replacing batteries is high,so the network lifetime is one of the most important design considers in Lo Ra networks.Existing work mainly uses physical layer methods to allow the Lo Ra receiver to resolve multiple conflicting Lo Ra signals,thereby improving the reliability and reducing the energy consumption.This thesis studies the resource allocation in the Lo Ra network to achieve the fairness of energy consumption among the devices,thereby improving the network lifetime.Furthermore,this thesis considers the impact of wireless link dynamics on energy fairness,and studies dynamic resource allocation to adapt to the dynamic Lo Ra network and improve the network lifetime.This work has been published in IEEE ICDCS2019 and IEEE ICNP 2020.(2)The Low Power Personal Area Networks are self-organizing low power networks composed of a group of Io T devices,and the Io T data is transmitted to the sink node and the remote server for analysis and processing in a multi-hop manner.Since the signal transmission distance of Io T devices is relatively short and the deployment is dense,the low power Io T networks allocate different channels for the devices to avoid conflicts between signals.In addition,the Low Power Personal Area Networks use Time Divided Multiple Access(TDMA)technology to divide the time into multiple time slots,and allocate time slots and wireless channel resources for different transmission links.Existing work often assumes that the wireless links will succeed in a time slot.However,the wireless signal is unreliable,and the lost packets can only be retransmitted in the next transmission period,which greatly increases the data transmission time.Under the premise of ensuring the deadline for data transmission,this thesis considers the unreliability of data transmission and the link interference under multi-hop network topology.Based on the urgency of different data links,the wireless channel resources and time slot resources are allocated to reduce interference in the network and improve the reliability of data transmission.This work has been published in IEEE Transactions on Industrial Informatics 2020,Pervasive and Mobile Computing 2017,and IEEE INFOCOM Workdshop Mise Net 2018.(3)Due to the limited computing resources and energy of low-power Io T devices,in order to cope with increasingly complex Io T tasks such as data and image analysis,lowpower Io T devices can offload their complex tasks to nearby edge servers.Edge servers can be base stations and wireless access points that nearby the low-power Io T devices.Multiple edge servers can be deployed to serve low-power Io T networks at the same time to improve data transmission and the efficiency of task processing.Since the Io T devices are wirelessly connected to the edge servers,as hardware resources,the allocation of edge server location will have a great impact on the performance of Io T data offloading.Existing research work mainly focuses on the location allocation of aggregation nodes in low-power personal area networks.However,when allocating the location of edge server resources,the heterogeneity of low-power Io T devices and the instability of wireless links Will affect the efficiency of data offloading and reduce the resource utilization of edge servers.To solve this problem,this thesis comprehensively considers the impact of heterogeneous low-power Io T devices and wireless link quality,and studies the resource allocation mechanism of edge servers to improve network throughput and resource utilization.This work has been published in IEEE Internet-of-Things Journal 2018.(4)Low-power Io T systems can execute complex computing tasks such as various machine learning algorithms on edge servers or cloud servers.Federated learning does not require Io T devices to upload collected data and perform model training locally,so it is widely used in data privacy-oriented Io T applications such as the Industrial Internet of Things.Existing research mainly focuses on reducing the training latency of federated learning in wireless mobile networks,and often overlooks the dense deployment of Io T devices and serious interference between wireless links in the Industrial Internet of Things,which may cause data packet loss and increase training latency.To solve the above problems,this thesis proposes the wireless resource and Io T devices allocation scheme,and comprehensively considers the influence of the above two resources on the training latency of federal learning,which greatly reduces the training delay of the global model.This work has been submitted to IEEE Transactions on Industrial Informatics 2021(under revision).