Deep Reinforcement Learning-based Partial Task Offloading in High Altitude Platform-aided Vehicular Networks
Tri‐Hai Nguyen, Thanh Phung Truong, Nhu–Ngoc Dao, Woongsoo Na, Heejae Park, Laihyuk Park
Abstract
Compared with traditional terrestrial access networks, a mobile edge computing-enabled aerial access network is a potential paradigm for performing complicated computations by offloading tasks to edge servers. To this end, high altitude platforms employed in the stratosphere to offer extensive coverage and powerful computing capabilities are regarded as a crucial component of an aerial access network. In this work, we investigate a computation offloading issue in a high altitude platform -aided vehicle network, where a ground base station is either overload or not accessible. Non-orthogonal multiple access is employed to enhance the transmission rate. We model the problem as a Markov decision process due to the complexity of the vehicular network and computation-intensive, latency-sensitive tasks. Then, we develop a deep reinforcement learning-based intelligent offloading strategy to minimize all vehicles' total energy consumption and task execution latency.