Latency Optimization in UAV-Assisted Mobile Edge Computing Empowered by Caching Mechanisms

Abstract

Mobile edge computing (MEC) revolutionizes data processing by shifting it from the network core to the edge, significantly reducing latency and ensuring Quality of Service. Integrating the agile and flexible unmanned- aerial-vehicle (UAV) technology with MEC offers new opportunities and challenges in decision making for dynamic and complex environments due to the UAVs’ mobility and Line of Sight advantages. Motivated by the potential of UAV-assisted MEC systems with caching mechanisms, this study addresses the optimization problem under uncertain conditions and user demand. To tackle the complex nonconvex sequential decision problem, a deep reinforcement learning framework named delay hybrid action actor-critic is proposed, possessing the capability to handle scenarios requiring both continuous and discrete actions. Comprehensive simulations are conducted to validate the capability of the proposed framework, demonstrating its superiority over traditional methods.