Research On Dynamic Resource Allocation Algorithm In Unmanned Aerial Vehicle Aided Edge Computing

The global deployment of 5G technology and the prospective planning of 6G networks provide a network environment with higher rate,lower latency,and broader coverage for the Internet of Things(Io T).Complex functionalities and services in Io T applications necessitates the collection of extensive on-site data.Through utilizing distributed computing resources of edge computing technology,the applications achieve low latency and high energy-efficient data processing.and thus enable the intelligent control and management of the physical world.However,the application of edge computing in Io T encounters certain limitations.Firstly,in Io T application scenarios where infrastructure is scarce or power supply is unstable,there is a lack of communication and computing resources.Secondly,the proliferation of 5G technology leads to a surge in the number of devices,making it difficult for traditional communication facilities to meet the growing demands for data transmission and processing.Therefore,facing the rapid expansion of Io T application scope and the continuous increase in complexity,relying solely on traditional infrastructure is insufficient to provide an adequate amount of communication and computing resources.With the gradual maturation of unmanned aerial vehicle(UAV)technology and the rapid development of aerial-ground integrated networks,UAVs provide viable solutions to these problems.Characterized by great mobility and flexibility,UAVs can carry micro base stations or edge computing servers to fly or hover over the network,providing high-quality data transmission and real-time computing services to ground devices.Currently,research on UAV-aided edge computing systems has yielded fruitful results,but there are still some pressing issues to be addressed.Firstly,existing studies mostly focuses on the impact of either UAVs or Io T devices’ attributes on the network,but has not designed data collection algorithms considering the requirements of both comprehensively.Secondly,existing research predominantly uses latency as the criterion for measuring the quality of computing services,overlooking the freshness of computing tasks for Io T devices,which may not meet the timeliness requirements of rapid response applications.Furthermore,current resource allocation algorithm designs generally assume deterministic channels for communication links,which does not align with the dynamics of UAV-aided edge computing systems and may not be applicable in real networks.Lastly,existing research typically assumes that ground devices can obtain the perfect state information of UAVs,which is rarely the case in practical applications.In summary,there are four technical challenges existing in UAV-aided edge computing systems: 1)How to design UAV data collection algorithms based on the inherent attributes of UAVs and Io T devices; 2)How to design dynamic resource allocation algorithms that ensure the timeliness of network information; 3)How to design effective dynamic resource allocation algorithms to improve system performance considering the dynamics of UAV-aided edge computing systems; 4)How to achieve task offloading and design corresponding resource allocation algorithms in the absence of perfect UAV state information for ground devices.To address these issues,this thesis progressively delves into the dynamic resource allocation algorithms in UAV-aided edge computing systems from four different perspectives:In the first part,focusing on time-constrained scenarios,the optimization algorithm for service sequence and hovering location in UAV-aided edge computing is studied.This thesis considers a more practical multi-UAV data collection system where UAVs with limited battery and storage capacities fly between locations and collect data from the corresponding ground Io T devices at these locations within time windows.The opening time of the time window indicates the earliest time at which the device can upload data,and the closing time represents the latest completion time of data uploading.Based on the characteristics of this system,this thesis improves the traditional ant colony algorithm and combines it with the successive convex approximation(SCA)method to design an service sequence and hovering location optimization algorithm.By optimizing the service sequences and hovering locations of UAVs,this algorithm effectively reduces the number and the operating time of UAVs while satisfying the battery and storage constraints of UAVs and the time window requirements of devices.In the second part,the age of information(Ao I)is utilized as an metric to measure the freshness of computing tasks offloaded by the Io T devices,and the joint resource allocation algorithm in the UAV-aided edge computing system is studied.This thesis introduces the concept of Ao I into the UAV-aided edge computing system,defined as the time elapsed from the generation of the latest executed task of the device to its completion.In this system,the computing services provided by multiple hovering UAVs to ground Io T devices need to meet the Ao I requirements of devices.Based on this system model,this thesis derives the average peak age of information for Io T devices and uses it as their service quality requirement.With the goal of reducing system energy consumption,a joint resource allocation and hovering location optimization algorithm is proposed.By optimizing the offloading strategies of devices,the hovering locations of UAVs,and communication and computing resource allocation,this algorithm reduces the transmission energy consumption of devices and the computing energy consumption of UAVs while meeting the Ao I requirements of devices.In the third part,under the premise of time-varying channels,the joint resource allocation algorithm in the UAV-vehicle cooperative edge computing system is studied.To address the problem of vehicles having difficulty providing high-quality services to devices far from the road,this thesis proposes a UAV-vehicle cooperative edge computing architecture.This architecture effectively improves the quality of computing services by leveraging the mobility and flexibility of onboard UAVs.However,the mobility of vehicles and UAVs results in time-varying channels between them and ground Io T devices.To address this issue,utilizing the first-order Gaussian-Markov process to estimate the channel state information,the lower bound of signal-to-noise ratio in the time-varying Rician channel is obtained.Based on this,a joint communication and computing resource allocation and trajectory planning algorithm is proposed,which increases the total computation quantity of the system,while ensuring that all data are executed before UAVs return to vehicles.In the fourth part,based on sensing technology,the joint resource allocation algorithm in the UAV-assisted vehicle-to-everything system is studied.Addressing scenarios involving the mobility of ground users(vehicles),a sensing-enabled UAV-aided vehicular network structure is proposed.In this structure,vehicles lack accurate state information of UAV before offloading computing tasks,and need to track and predict the state based on radar sensing technology.Leveraging the mobility characteristics of vehicles and UAV,this thesis derives state evolution and radar measurement models,and designs a UAV state prediction and sensing model by utilizing extended Kalman filtering.Based on this model,a joint communication and computing resource allocation algorithm is proposed.By jointly optimizing the allocation of time slot,transmit power,and computing resource,this algorithm reduces system energy consumption while ensuring the completion of tasks for each vehicle.