RF neuromorphic spiking sensor for smart IoT devices

Abstract

For a ubiquitous sensing in Internet of Things (IoT), a large number of low-cost devices with ultra-low power massive communications are required. Heavy cloud computing pressure and low-intelligent traditional front-end hardware are major challenges for extending sensing in IoT applications. While state-of-the-art is focusing on the cloud-edge platform solutions and the optimization of the performed and transmitted data, this paper proposes an intelligent equipment hardware for smart IoT devices. A radiofrequency (RF) neuromorphic spiking sensor is implemented in BiCMOS 55 nm technology, comprising a wake-up receiver with a spiking pre-processing neural system (artificial synapses and neurons). The proposed system can identify bit patterns of two signals modulated using On–Off Keying at a frequency of 2.4 GHz received from two IoT receivers. Moreover, the system can recognize the orientation of the mobile IoT transmitter. This can be achieved based on the output spiking frequency of the neuron responsiveness over the difference between the input powers. Post-layout simulations demonstrate that the orientation of the source can be detected for various distances between the source and the two receivers. Significant performances are obtained with 1.1 nW of total power consumption and 0.7 fJ/conv of energy efficiency.