Al-Driven Optimization in UAV-Assisted Edge-Cloud Systems: DRL and MADRL Approaches

The rapid growth of Internet of Things (IoT) applications and latency-sensitive services has posed significant challenges to traditional centralized cloud computing architectures. Edge-cloud computing has emerged as a promising solution by distributing computational resources closer to data sources. However, the resulting distributed and highly dynamic environments, particularly mobile edge computing systems assisted by unmanned aerial vehicles (UAVs) introduce complex optimization challenges in task offloading, resource allocation, and trajectory planning. Classical optimization methods, which rely on static models and precise system assumptions, often fail to scale or adapt effectively in such environments. Recently, deep reinforcement learning (DRL) and, more notably, multi-agent deep reinforcement learning (MADRL) have gained significant attention as data-driven and adaptive solutions for decision-making in complex distributed systems. This paper presents a systematic and comprehensive review of state-of-the-art research published between ۲۰۲۳ and ۲۰۲۵, focusing on the application of DRL and MADRL for optimization in distributed edge-cloud systems. Special emphasis is placed on UAV-assisted mobile edge computing scenarios, including joint task offloading, resource management, and trajectory optimization. The reviewed studies indicate that multi-agent learning architectures often offer superior scalability and coordination in heterogeneous and dynamic environments compared to single-agent approaches. This enables more effective trade-offs among conflicting objectives, such as latency, energy consumption, and system stability. Finally, the paper discusses open challenges and promising future research directions, including federated reinforcement learning, scalability, and security-aware optimization.

کلیدواژه ها:

Deep Reinforcement Learning ، Multi-Agent Reinforcement Learning ، Distributed Systems Optimization ، Mobile Edge Computing (MEC) ، UAV-Assisted Edge Computing ، Task Offloading