A Dynamic DAG-Based Scheduling Strategy for Cross-Chain Transaction Management

Abstract

Cross-chain sharding improves interoperability and scalability, yet relay-chain execution remains constrained by dynamic dependencies and shard contention. This paper characterizes the relay-chain execution mechanism of cross-chain sharded transactions and models registration, recording, and confirmation transactions as a dynamic DAG, so that scheduling decisions respect the prerequisite dependencies required by the underlying cross-chain atomicity and consistency mechanisms. On this basis, we propose Fast Out-degree based Monte Carlo Tree Search (FO-MCTS), a dynamic DAG scheduling algorithm for cross-chain sharding. FO-MCTS uses dependency relationships to guide executable-node priority and shard-search expansion, enabling dependency-continuous scheduling while avoiding exhaustive shard-assignment search. Simulation results show that FO-MCTS improves transaction throughput by about 25% on average, reduces the cross-shard communication ratio by 65%–85%, and maintains lower scheduling-induced delay.