Cognitive computation and communication: A complement solution to cloud for IoT

Abstract

The Internet of Thing (IoT) is experiencing explosive growth in the number of devices and applications. However, the existing cloud-centric architecture of IoT poses serious challenges regarding network latency, privacy, and energy-efficiency. We have presented COGNICOM+ concept, a brain-inspired software-hardware paradigm, to support IoT's future growth and developed 4 research directions - flexible radio, convolutional neural network accelerator, compressed deep learning, and game theory for reasoning and collaboration - within COGNICOM+. The key idea is to bring computing closer to the end-user while focusing on optimal uses of local smart application gateway and cloud computing. COGNICOM+ consists of two key components: Cognitive Engine (CE) and Smart Connectivity (SC). The cognitive engine is powered by deep-learning algorithms integrated with game-theoretic decision analytics, implemented on a low-power application-specific integrated circuit. It provides cognitive functions to smart objects. The smart connectivity integrates neural network inspired designs of cognitive radio, transceivers, and baseband processors. The SC provides flexible and reliable connections to IoT objects and optimally distributes communication resources.