BRL-Net: A Blockchain-Based Task Offloading Framework Using Smart Contracts for Metaverse

Abstract

The emergence of the Metaverse has introduced significant challenges in task offloading and data processing due to its virtual universe nature with immersive environments and a multitude of interconnected users and devices. The abundance of data in the Metaverse poses security challenges in local processing, necessitating traditional methods such as data transfer to Mobile Edge Computing (MEC) and subsequently to the cloud, thereby emphasizing security concerns. In this paper, a novel approach to address these challenges has been introduced: An Ethereum Blockchain-based MEC framework uses smart contracts designed to ensure secure task offloading. It enables authentication in the Metaverse through smart contracts, followed by modeling the task offloading issue as a Markov Decision Process (MDP). To solve this MDP problem, a hybrid algorithm integrating Deep Q-Networks (DQN) with Bidirectional Long Short-Term Memory (Bi-LSTM), known as BRL-Net (Bi-LSTM Reinforcement Learning Network), has been proposed. This framework enables secure and efficient task offloading in dynamic Metaverse environments. BRL-Net outperforms Proximal Policy Optimization (PPO), achieving a 9.93% higher reward and greater stability. The BRL-Net's performance across Blockchain consensus mechanisms shows Delegated Proof of Stake (DPoS) as the most efficient, reducing latency by 49.96%, increasing throughput by 10.48%, and lowering energy consumption by 50.24%, compared to Proof of Stake (PoS), thereby optimizing Metaverse performance.