Sustainable broadcasting in Blockchain Networks with Reinforcement Learning
Danila Valko and Daniel Kudenko
BUIS-Tage 2025 – Smarte und Nachhaltige Infrastrukturen
Bitcoin and Ethereum generate an estimated 64 and 26 million tons of CO2 annually. To reduce blockchain energy consumption, various strategies have been explored, including alternative consensus mechanisms and redundancy reduction. This paper focuses on the latter by introducing a reinforcement learning (RL)-based method to optimize block broadcasting. Our RL agent reprioritizes block propagation using real-time network data, reducing both message volume and propagation time, which contributes to lower energy use across distributed networks. We extended a blockchain simulator to integrate with the RL framework, and simulations confirmed our approach outperforms the default scheme. Our setup also supports future research into ML-enhanced blockchain protocols.
June / 2025
conference
OFFIS Autoren
