APRES: Adaptive Partial Reconfiguration for Blockchain Sharding via Deep Reinforcement Learning
Concurrency and Computation: Practice and Experience · 2026-08 · 38 (15)
DOI: 10.1002/cpe.70883 ↗Kurz gefasst
Fachartikel in der Zeitschrift Concurrency and Computation: Practice and Experience, online erschienen am 29.07.2026. Verfasst von Shihui Lin und 2 weiteren Autor:innen. Die Arbeit nennt 57 Literaturangaben. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Sharding improves blockchain scalability by partitioning the network into smaller shards that process transactions in parallel. Shard configurations degrade over time as transaction workloads shift and adversarial validators may accumulate within individual shards. Periodic reconfiguration is therefore necessary, but reshuffling all shards at every epoch incurs heavy state migration and cross‐shard coordination overhead. We propose APRES, a deep reinforcement learning framework for partial shard reconfiguration that selectively determines which shards to reconfigure at each epoch boundary. APRES adopts a two‐tier architecture with an agent shard and multiple transaction shards that execute consensus. We implement APRES on a sharded blockchain testbed deployed on Alibaba Cloud and evaluate it using real Ethereum transaction data. Experimental results show that compared to fixed‐period baselines, APRES reduces reconfiguration frequency by 35% and improves throughput by 18% without degrading system security.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Lin, S., Huang, H., & Cheng, Y. (2026). APRES: Adaptive Partial Reconfiguration for Blockchain Sharding via Deep Reinforcement Learning. Concurrency and Computation: Practice and Experience, 38(15). https://doi.org/10.1002/cpe.70883
Vancouver
Lin S, Huang H, Cheng Y. APRES: Adaptive Partial Reconfiguration for Blockchain Sharding via Deep Reinforcement Learning. Concurrency and Computation: Practice and Experience. 2026;38(15). doi:10.1002/cpe.70883
BibTeX
@article{Lin2026,
author = {Lin, Shihui and Huang, Hui and Cheng, Yaohong},
title = {APRES: Adaptive Partial Reconfiguration for Blockchain Sharding via Deep Reinforcement Learning},
journal = {Concurrency and Computation: Practice and Experience},
year = {2026},
volume = {38},
number = {15},
doi = {10.1002/cpe.70883}
}
EUVIMED