Partial Offloading MEC Optimization Scheme using Deep Reinforcement Learning for XR Real-Time M&S Devices

With the advent of 5G, the development of extended reality (XR) technology, which combines augmented reality (AR), virtual reality (VR), and advanced human-computer interaction (HCI) technology, is considered one of the key technologies of future metaverse engineering. Especially, XR real-time modeling and simulation (M&S) devices that can be applied to various fields (e.g., emergency training simulations, etc.) have tasks with large amounts of data to be processed. However, if the XR task is processed only by wireless user equipment (UE), the UE's energy may be quickly depleted, and the quality of service (QoS) may not be satisfied. To solve these problems, this paper proposes a partial offloading optimization scheme through multiple access edge computing (MEC). In addition, deep reinforcement learning (DRL) is used to reflect the dynamic state of the MEC system and to minimize the delay. The simulation results show that the proposed scheme optimizes the delay performance by efficiently offloading the XR tasks.