Offloading among multiple radio access technologies (RATs) is an effective solution to tackle some of the key challenges faced by 5G wireless networks. While making an offloading decision, in user-centric offloading schemes, the network context is not taken into account, which limits the full utilization of the potential offered by these techniques. Therefore, in this work, we develop a data offloading framework wherein a user defines a tuple, which includes a request for data service and the associated delay limit, and the user equipment (UE) translates this tuple into an optimal offloading policy by taking into account the network context. The data offloading problem is formulated by assuming a finite horizon Markov decision process (MDP), and an analytical data offloading model is derived by using stochastic geometry for modeling the spatial domain of a multi-RAT wireless network, and by employing Markov process for capturing mobility of users. In order to validate the performance of derived model, we also solve the offloading problem by using policy iteration (PI) algorithm. The results show that the analytical algorithm outperforms the iterative algorithm. We also benchmark the proposed framework against standard data offloading methods.

中文翻译：

---

面向多 RAT 5G 无线网络的自动数据卸载框架

多种无线接入技术 (RAT) 之间的卸载是解决 5G 无线网络面临的一些关键挑战的有效解决方案。在做出卸载决策时，在以用户为中心的卸载方案中，没有考虑网络上下文，这限制了这些技术提供的潜力的充分利用。因此，在这项工作中，我们开发了一个数据卸载框架，其中用户定义一个元组，其中包括对数据服务的请求和相关的延迟限制，用户设备（UE）通过考虑将这个元组转换为最佳卸载策略考虑网络上下文。数据卸载问题是通过假设有限范围马尔可夫决策过程 (MDP) 来制定的，并且通过使用随机几何对多RAT无线网络的空间域建模，并通过采用马尔可夫过程来捕获用户的移动性，导出分析数据卸载模型。为了验证派生模型的性能，我们还使用策略迭代（PI）算法解决了卸载问题。结果表明，解析算法优于迭代算法。我们还针对标准数据卸载方法对提议的框架进行了基准测试。