IoT is nothing but inter-connection of devices (fitted with sensors) with one another, over the Internet. However this architecture of IoT is limited to certain horizons which is overcome by the cognition aspect which when integrated with IoT leads to the development of Cognitive IoT.
Cognitive IoT deals with following three aspects :-
- Understanding
- Reasoning
- Learning
Due to the above aspects, IoT gets to behave more like a human aware system rather than being only an interconnectivity mechanism for devices to exchange data and cater the user needs. The CIoT (Cognitive IoT) still caters the need of users but those needs are now intelligently catered such that the user is satisfied as well as minimal resources, like channel usage, are used for catering the user needs.
Now the need to cluster heterogeneous devices in CIoT arises inorder to study correlations and patterns across different unstructured data and group those devices together which can facilitate better data transmission due to lesser number of data packets being send for the common queries which are raised by the devices within a cluster. Thus this reduces the data traffic and provides faster communications.
The goal of this project is to develop an efficient algorithm which is capable of dynamically clustering the devices based on correlations between data and the devices producing the data. The data can be of any dimension and unknown to the system initially therefore a Cognitive layer of IoT is considered for understanding and learning about the completely new and unknown data. Moreover, to reduce data volume being send from each cluster, a further modification would be added to the algorithm via consideration of development of cluster heads within each cluster.