Comprehensive survey on reinforcement learning-based task offloading techniques in aerial edge computing

Abstract

Aerial edge computing (AEC) has emerged as a pivotal platform offering low-latency computation services, seamless deployability, rapid operationality, and high maneuverability to the Internet of things (IoT) devices of end users. Different aerial computing platforms offer different computation support to process the tasks of IoT devices, which affects the offloading decision. However, effective task offloading (TO) decision-making in this context remains a critical challenge because it impacts the quality of service, energy consumption, resource allocation, and latency requirements. Most current research uses reinforcement learning (RL)-based offloading decisions in AEC owing to the uncertainty of the environment and the heterogeneity of computation platforms. Therefore, this survey explores the prevailing use of RL-based algorithms for TO in AEC, addressing the inherent uncertainty of the environment and the heterogeneity of computation platforms. This study systematically reviews and compares RL-based techniques employed for efficient offloading decisions in heterogeneous aerial computing platforms. It delves into recent research findings, highlighting the various approaches and methodologies applied. Additionally, the paper provides a comprehensive overview of the performance metrics widely used to evaluate the efficacy of RL-based offloading decision techniques. In conclusion, this survey identifies research gaps and outlines future directions, aiming to guide scholars and practitioners in advancing the field of RL-based TO in AEC.