A Caching-Enabled Permissioned Blockchain Framework for Industrial Internet of Things based on Deep Reinforcement Learning

Abstract

The integration of Industrial Internet of Things (IIoT) and blockchain has become a popular concept. Numerous IIoT nodes together form a decentralized network with rich location-aware computation resources, which can provide great data processing capability and offer low-latency services. However, we are still facing the challenges of how to efficiently process the massive IIoT data on resource-constrained IIoT nodes by blockchain smart contracts, as their storage only allows them to store limited blockchain data. This work aims to improve the smart contract execution efficiency on these IIoT nodes by caching based on Deep Reinforcement Learning. On the one hand, focusing on the characteristics of IIoT, the ledger structure, network architecture, and transaction flow are optimized. IIoT nodes are enabled to cache part of block data without affecting global data consistency. On the other hand, we formulated the blockchain caching problem as a Markov Decision Process and implemented a lightweight caching agent based on deep Q-learning. Proper features and a reward function are defined to minimize the execution delay of smart contracts. The extensive experiment results show that our proposed framework can effectively reduce the block dissemination costs and the agent can provide satisfactory caching performance.