AI-Driven Resource Allocation in Ethereum Blockchain NetworksUsing Hybrid Q-Learning and Ship Rescue Optimization

Abstract

Resource allocation in blockchain networks is an urgent issue due to changes in transaction load, networkcongestion, and the computational challenges associated with smart contract execution. Suboptimal resource utilization leads to high operational costs and reduced network performance. In this context, this paper presents a new hybrid algorithm for resource management in blockchain networks based on the integration of Q-learning reinforcement learning and the Ship Rescue Optimization (SRO) algorithm. The SRO algorithm is used to optimize the hyperparameters and the initial Q-learning policy, enabling the learning process to converge more effectively to an optimal solution and make better resource allocation decisions. We formulate the resource allocation problem as a Markov Decision Process (MDP), in which the agent learns optimal scaling policies for CPU, memory, and bandwidth resources. Comprehensive testing on an implemented blockchain network with 100 nodes and 245,782 transactions across 1,000 blocks shows significant improvements, including an average reduction of 38.76 % in CPU usage, 35.99% in RAM usage, and 38.51% in transaction latency, along with a 58.21% increase in throughput compared to baseline methods. All improvements are statistically significant (p 3.0) according to the statistical analysis. The ablation study shows that each component plays a significant role in the overall system performance. In particular, the proposed hybrid algorithm provides an additional 13.88% performance improvement compared to Q-learning alone and . Furthermore, the suggested framework outperforms existing methods, such as Deep Q-Networks, Genetic Algorithms, and Particle Swarm Optimization, establishing a new benchmark for blockchain resource allocation.