Sharding technology aims to alleviate the performance and scalability limitations of blockchain protocols by distributing transaction (TX) processing across multiple node clusters. In the context of state sharding, cross-shard account migration is essential for minimizing cross-shard transactions and achieving balanced workload distribution. However, most existing migration schemes rely on locking mechanisms to ensure transaction isolation, which often leads to reduced system throughput and increased risk of deadlocks. To overcome these limitations, we propose a lock-free scheduling (LFS) approach, which enables the coordination of migration transactions without introducing locks. LFS allows migrating accounts to function as both payers and payees during migration, thereby improving concurrency and enhancing overall throughput. The core idea involves decomposing each transaction into sub-transactions that can be processed in parallel across multiple shards. Experimental results show that LFS consistently outperforms baseline schemes under lock-free conditions, reducing transaction latency by 13% and improving throughput by over 8%.