Recent years have witnessed a blooming of new applications that demand different network services. Network slicing is advocated by the research community to simultaneously support multiple services on a common physical infrastructure. Federated network slicing, which involves multiple operators, further generalizes the concept to cover a broader range. Existing federated slicing systems advocate the master-slave architecture among untrusted operators, which brings some centralization concern, making operators hesitate to join the system. Recently, blockchain shows great power to build trust in decentralized environments. Besides, artificial intelligence (Ai), especially reinforcement learning, is envisioned with the potential to develop more efficient optimization algorithms. Motivated by innovations in blockchain, smart contract, and Ai, this article proposes a decentralized federated slicing architecture that is trustful and efficient. We systematically discuss the design principles and key challenges in realizing the blockchain-enabled architecture. With these principles and challenges in mind, we develop a general architecture for multiple operators and cloud providers, with a new proof of business consensus protocol to ensure incentive and fairness. To further enhance its efficiency, we utilize reinforcement learning to accelerate optimizations in the resource allocation. Benefits of the Ai accelerated optimizer are demonstrated in simulations.

中文翻译：

---

具有AI加速优化功能的5G网络的区块链支持联合切片

近年来，见证了需要不同网络服务的新应用程序的兴起。研究机构提倡网络切片，以在一个通用的物理基础架构上同时支持多种服务。涉及多个运营商的联合网络切片进一步将这一概念推广到更广泛的范围。现有的联合切片系统在不受信任的运营商中倡导主从架构，这带来了一些集中化问题，使运营商不愿加入该系统。最近，区块链显示了在分散环境中建立信任的强大能力。此外，人工智能（Ai），尤其是强化学习，有望开发出更有效的优化算法。受区块链，智能合约和Ai创新的推动，本文提出了一种可信且高效的分散式联合切片架构。我们系统地讨论了实现区块链架构的设计原则和主要挑战。考虑到这些原则和挑战，我们为多个运营商和云提供商开发了通用架构，并提供了新的业务共识协议证明以确保激励和公平。为了进一步提高其效率，我们利用强化学习来加速资源分配的优化。仿真显示了Ai加速优化器的优势。考虑到这些原则和挑战，我们为多个运营商和云提供商开发了通用架构，并提供了新的业务共识协议证明以确保激励和公平。为了进一步提高其效率，我们利用强化学习来加速资源分配的优化。仿真中展示了Ai加速优化器的优势。考虑到这些原则和挑战，我们为多个运营商和云提供商开发了通用架构，并提供了新的业务共识协议证明以确保激励和公平。为了进一步提高其效率，我们利用强化学习来加速资源分配的优化。仿真显示了Ai加速优化器的优势。