Adaptive Task Offloading for Mobile Aware Applications Based on Deep Reinforcement Learning

With the widespread popularity of mobile devices and wireless communication technology, people's demand for the services that mobile devices can provide is also generally increased. However, the computing capacity of the mobile device itself is extremely limited and cannot be adapted to some computation-intensive tasks. This is why mobile edge computing (MEC) comes in. One of the problems that must be faced is how to adaptively offload computing tasks between changing computing tasks and dynamic wireless network bandwidth. In this paper, we propose an adaptive offloading strategy based on the deep reinforcement learning algorithm, which can generate a more reasonable scheduling strategy for different task flow structures and multi-thread data offloading. Our simulation results show that there is up to 20% improvement in user experience compared to traditional offloading strategies on average.