Recently, one of the most promising architecture for the Internet of Things (IoT) is Information-Centric Networking (ICN). The use of data caching techniques reduces the total energy consumption of the system and the response time. Many IoT devices are resource-constrained in terms of energy and storage. On the other hand, IoT data are usually transient, and their lifetime is short. Additionally, the popularity of IoT data varies with time. Therefore, static cache policies are not appropriate for these dynamic environments. In this paper, a sliding window-based adaptive cache placement framework is proposed for ICN-IoT architecture. An optimization model is developed based on integer linear programming to minimize the total energy consumption of the system according to device constraints and IoT data attributes, including freshness and content popularity. A software-defined cache controller is designed in both centralized and decentralized modes to make optimal cache placement policy with the capability to adapt to the changes of data popularity over time. At the beginning of each time slot, the popularity values are estimated by a neural network predictor and applied to the optimization model. This predictor benefits from online learning, and it is adapted to the newly observed data at the end of each time slot. The proposed approach is compared with some of the conventional caching methods. The results demonstrate a significant reduction in total energy consumption and robustness to the sudden changes of data popularity.

Keywords