An LSTM-Method-Based Availability Prediction for Optimized Offloading in Mobile Edges

Abstract

Mobile edge computing (MEC) can augment the computation capabilities of a vehicle terminal (VT) through offloading the computational tasks from the VT to the mobile edge computing-enabled base station (MEC-BS) covering them. However, due to the limited mobility of the vehicle and the capacity of the MEC-BS, the connection between the vehicle and the MEC-BS may be intermittent. If we can expect the availability of MEC-BS through cognitive computing, we can significantly improve the performance in a mobile environment. Based on this idea, we propose a offloading optimization algorithm based on availability prediction. We examine the admission control decision of MEC-BS and the mobility problem, in which we improve the accuracy of availability prediction based on Empirical Mode Decomposition(EMD) and LSTM in deep learning. Firstly, we calculate the availability of MEC, completion time, and energy consumption together to minimize the overall cost. Then, we use a game method to obtain the optimal offloading decision. Finally, the experimental results show that the algorithm can save energy and shorten the completion time more effectively than other existing algorithms in the mobile environment.