Edge device resource optimization (EDRO): evaluating the impact of reinforcement learning and deep reinforcement learning methods

In the realm of edge computing, an extension of cloud capabilities to the network periphery, resource allocation presents intricate challenges. This complexity is a result of the dynamic nature of the environment, characterized by a plethora of devices, intermittent data flows, and diverse applications. Uneven resource distribution can lead to operational issues and task failures, adversely affecting system efficiency. To address these challenges, we introduce an innovative framework that harnesses and compares the power of Reinforcement Learning (RL) and Deep Reinforcement Learning (DRL) for resource allocation. Our methodology aims to optimize computational container status and allocate resources. DRL enables us to internalize optimal resource allocation strategies based on real-time contextual cues. Through the lens of DRL, our advanced approach adapts to the inherent volatility of edge computing environments, resulting in judicious resource allocation decisions that enhance edge node performance. This study provides a comparative analysis of Reinforcement Learning (RL) and Deep Reinforcement Learning (DRL) in the context of resource allocation for edge computing. By employing DRL, our framework demonstrates the capability to outperform traditional RL methods, adapting more effectively to the dynamic nature of edge environments and yielding superior resource allocation outcomes.