Deep learning based self-adaptive framework for environmental interoperability in internet of things

IoT interconnects various entities including users, devices, information, and services, thus, interoperability is essential to realize the Internet of Things (IoT). There are various perspectives to support interoperability IoT environment, and one interoperability problem is related to constructing IoT environments at runtime. The problem is caused by that it is hard to predict IoT environments at design time. In other words, the IoT environment can be dynamically changed, thus an IoT system has to adapt to the change. To solve the environmental interoperability, a self-adaptive framework based on deep neural networks (DNN) is proposed to construct IoT systems at runtime. The proposed framework provides requirement verification and adaptation at runtime. Arduino-based IoT environments were implemented, and experiments were performed to show the efficiency. The results showed the reasonable performance to verify requirement satisfaction using DNNs.