Machine Learning Based Edge-Assisted UAV Computation Offloading for Data Analyzing

Recently, Combining communication technology with Unmanned Aerial Vehicle (UAV) have been regarding as one of the promising techniques in the future network. Various services can be served through UAV, such as information collection in disaster area or traffic monitoring in smart city. However, on-board computation resource and the battery lifetime of UAV are limited. For that reason, there is a limit to perform analysis, such as, machine learning using collected data. In this situation, UAV can get help from an adjacent Mobile Edge Computing (MEC) server that has computing resources. However, the closest MEC server does not always guarantee optimal computing performance and communication efficiency. In this paper, to solve this problem, we propose the overall system of machine learning based UAV to MEC computation offloading for data analyzing. First, we define the problems into two. Firstly, we define two problems 1) matching UAV and task cluster in consideration of energy efficiency. 2) Finding optimal MEC server to offload the task to minimize the total processing time and energy consumption. Then, we apply a machine-learning algorithm to solve the two problems. In the simulation section, we analyze the proposed method and greedy method in terms of total energy consumption and total processing time. Simulation results demonstrate that our proposed mechanism is efficient in total energy consumption of UAV and total processing time of tasks.