UAV-Aided Fair Data Collection And Mobile Edge Computing For Iot Based On Path Planning

With the development of 5G and Internet of Things(IoT)technology,the explosive growth of data volume of massive IoT devices has generated higher communication demand.However,the special working environment of some IoT devices brings difficulties to traditional cloud computing and stationary communication devices.In remote environments such as mountainous areas,wireless multi-hop communication cannot provide stable communication links for Internet of Things devices with limited communication capabilities,and redundant data transmission brings a lot of extra energy consumption.In addition,it is extremely costly to build stationary communications equipment that relies on power lines for temporary demand in complex and hazardous environments.As a result,more attention has been paid to communication using unmanned aerial vehicles(UAVs),which are flexible and low cost.By installing wireless communication modules and computing equipment on the UAV,the UAV can provide better communication coverage,data transmission and even auxiliary computing services for ground nodes by taking advantage of its high maneuverability and altitude advantages.The high maneuverability of UAV makes the dynamic flight path planning an important problem in UAV communication.This paper studies the flight path planning of UAV based on two kinds of problems,data acquisition and auxiliary computing.1)Aiming at the data acquisition problem in wireless sensor network(WSN),the paper establishes a communication model between UAV and ground sensor node based on typical air-to-ground channel.By abstracting the flight movement and data acquisition process of UAV into a Markov decision process(MDP),fairness-aware proximal policy optimization(FAPPO)algorithm,a deep reinforcement learning(DRL)algorithm for UAV flight path planning,is designed to maximize the fair perceived return.By comparing the simulation,the paper reveals the ground distribution of sensor nodes,sensor size of storage space,communication condition difference,the influence of FAPPO algorithm is verified and compared with other comparison algorithm is more efficient and fair data acquisition performance,and for the next chapter the user side of the auxiliary calculation of energy consumption optimization provides some transcendental conclusion and the work mentality.2)Aiming at the mobile computing problem of the computationintensive intelligent terminal network,the paper studies the path planning and task computing strategy optimization aiming at minimizing the userside energy consumption in the complex environment with larger state space and action space,according to the user’s characteristics of mobility and sensitivity to energy consumption.Aiming at the scenario of multiple UAVs,a multi-agent reinforcement learning algorithm named GAKMADDPG combined with the initial deployment of UAVs is proposed based on the channel model in the previous chapter.Specifically,the initial deployment positions of all UAVs are optimized by GAK algorithm,a clustering algorithm with the ability to seek global optimization,and then the movement and interaction between UAVs and intelligent terminals are modeled as Markov processes.Multi-agent reinforcement learning(MARL)algorithm MADDPG,which is good at dealing with high dimensional state action space,is used to solve the problem of trajectory planning and task calculation strategy optimization in dynamic environment.The simulation results show that the GAK-MADDPG algorithm can make full use of the computing ability of UAV and intelligent terminal by reasonably planning the trajectory and task calculation strategy of UAV,thus saving the energy consumption on the user side to the greatest extent and improving the user side experience.