Research On Joint Offloading And Trajectory Design In Uav Based Mec Networks For Iot Devices

With the continuous upgrading of mobile communication technology,the Internet of Things(IoT)has achieved rapid development.At present,More and more Internet of Things devices(IoTDs)have been used in our daily lives,such as traffic,smart home,wearable and other monitoring devices.However,computation resources limitation is still the challenge for IoTDs in communication systems.To address the computation problem,mobile edge computing(MEC)has been considered as an efficient solution to provide stable and low-latency services by deploying computing servers on the edge of IoT.However,deploying an IoT system that combines MEC still faces many challenges,such as the high-cost to install the entire infrastructure on the edge of IoT in the case of sparse distribution of IoTDs.At the same time,due to the flexibility,low-cost and maneuverability of unmanned aerial vehicle(UAV),applying UAVs with computing capabilities on the edge of IoT has been considered as an efficient solution to solve the above problems.However,deploying a UAV-assisted IoT system that combines MEC faces many challenges such as trajectory design and task calculation mode selection.Therefore,it is of great significance to study the joint design of computing task offloading and trajectory of UAV in the IoT architecture that combines MEC and UAV.This thesis mainly studies the scheduling optimization strategies in aUAV-based MEC system for IoTDs.Based on the optimization requirements of different network indicators,comprehensively considering the trajectory of UAV and the task calculation mode of each IoTD,the in-depth exploration is carried out.The main contributions can be summarized as follows:Firstly,for the IoT architecture that combines MEC and UAV,this thesis studies the scheduling optimization problem aiming at minimizing the total delay of the system.According to the characteristics of MEC and UAV,the communication model of the UAV-MEC system is reasonably constructed,and the optimization problem of minimizing the total delay of the system is modeled under the constraints that the UAV provides services for all IoTDs in the system and the task calculation mode of each IoTD should be local calculation or offloading calculation,etc.This problem has time-varying variables and a large number of variable combination modes,and it is difficult to find the optimal solution.To solve this problem,this thesis proposes a Q-learning-based reinforcement learning algorithm and proves the convergence of this algorithm.Simulation results show that the algorithm proposed in this thesis can significantly reduce the total delay of the system compared with the five benchmark algorithms,and the gap between the proposed algorithm and the global optimal solution is less than 2.8%.Moreover,the algorithm can dynamically decide the trajectory of UAV and the task calculation mode of each IoTD,and can effectively use the resources of the IoT architecture.Secondly,for the IoT architecture that combines MEC and UAV,the work in the first part focus on the scenario where the UAV provides services for all IoTDs in the system,however,the system will face the limitation of the total delay of the system in an emergency.In this case,the UAV may not be able to provide services for all IoTDs in the system.Therefore,this thesis studies the scheduling optimization problem aiming at maximizing the number of IoTDs served by the UAV in the system.The communication model of the UAV-MEC system is the same as the model in the first part.The reasonable explanation of the waiting delay of each IoTD and the satisfaction index function of IoTDs are presented in this thesis.Furthermore,the optimization problem of maximizing the number of satisfied IoTDs in the system is formulated under the constraints of total delay and various constraints.Since this problem still has time-varying variables and a large number of possible combinations,this thesis proposes a Q-learning-based reinforcement learning algorithm to solve the optimization problem and proves the convergence of the algorithm.Simulation results show that the algorithm proposed in this thesis can significantly increase the number of satisfied IoTDs in the system compared with the benchmark algorithms.Meanwhile,the proposed algorithm can effectively use the computing power of the local server and the MEC server,and can effectively help the IoT system to complete emergency tasks.