A Review of Dynamic Sharding in Blockchain with Deep Reinforcement Learning: Challenges and Emerging Trends

Blockchain networks face significant scalability and efficiency challenges as transaction volumes increase. Sharding, a partitioning technique that divides the blockchain into smaller, parallel-processing units, offers a promising solution. However, traditional static sharding lacks adaptability, leading to inefficiencies in resource utilization and security vulnerabilities. This paper critically reviews dynamic sharding, with a focus on Deep Reinforcement Learning (DRL) integration to optimize shard allocation. DRL techniques enable adaptive resource allocation, reduce cross-shard communication overhead, and improve transaction throughput. Advanced methods further enhance performance, as demonstrated in applications like UAV networks and IoT blockchain systems. Nevertheless, DRL-based sharding faces substantial challenges. This review synthesizes these challenges and emerging AI-driven trends, highlighting the need for further research. The findings provide a foundation for future research to address critical gaps in intelligent blockchain optimization.