With the rapid development of the Internet of Things, service paradigms are shifting from centralized to distributed collaboration to provide ubiquitous IoT services. However, the heterogeneous and unevenly distributed resources in IoT environments face challenges in ensuring service quality due to the lack of mutual trust among multiple entities. Integrating blockchain with IoT to establish a trustworthy collaborative mechanism has emerged as a promising trend. Nevertheless, the inherent contradiction between blockchain’s high computational overhead and the resource constraints of IoT devices remains significant. Moreover, most existing sharding techniques neglect the heterogeneity of IoT environments, making them insufficient for large-scale collaborative applications. To address these challenges, this paper proposes a multi-shard collaborative architecture tailored for resource-constrained IoT environments. The architecture first introduces a multi-dimensional dynamic reputation evaluation model to accurately assess node performance. On this basis, a reputation-aware dual-gateway architecture is designed, wherein each shard dynamically elects an intra-shard gateway node based on reputation scores, while a verifiable random function is employed to securely elect a global gateway node. Furthermore, a Merkle Patricia Tree-based smart contract service directory is constructed to enable efficient cross-shard service discovery. In addition, this paper designs a complete cross-shard smart contract invocation process and employs a Byzantine Fault Tolerance consensus mechanism with threshold signatures to ensure the security and reliability of cross-shard collaboration results. Simulation results demonstrate that, compared with traditional sharding schemes, the proposed architecture increases system throughput to nearly 1700 transactions per second, achieving approximately 2× performance improvement, while reducing transaction latency to around 50 ms.
