Service chaining offloading decision in the EdgeAI: A deep reinforcement learning approach

Many mission critical devices are increasing with upcoming 5G network to fulfill a low latency for a real time network service on smart factory, autonomous vehicle, etc. Distributed cloud computing system also has a key role to execute the various mobile devices, because, an edge computing is the nearest from the mobile devices to provide low latency and computation energy consumption. In this paper, we consider the autonomous vehicles with video live streaming services. Especially, the vehicles require a low transmission delay as within 10 ms. To reduce a latency with low energy consumption, we propose a service chaining offloading decision with a deep reinforcement learning. We split tasks of the vehicle per service function blocks which have their own role. So it can do partial offloading and user association in a On-Device Edge of the vehicle and in the SBS at the same time. We can get results that service chaining offloading decision gives more optimal energy consumption with low-latency to autonomous vehicle users.