세미나

 Although the keyword “Internet of Things” has already gained huge popularity, many of the early IoT devices involved a mere addition of internet connectivity to existing electronics and their potential was limited by the form factor of the original electronics. To overcome this limitation, we need a new platform for IoT devices in a small and mobile form factor. Reduced size and cost will increase the number of IoT devices and their abilities. Autonomous control and data processing will relieve the burden to an end-user for management and maintenance, and will ultimately allow the end-user to “install and forget” IoT devices. There are still many challenges towards this vision. Non-digital blocks such as power converters, sensors, and wireless transceivers do not simply follow the scaling trends of digital circuits and require more effort to reduce the area and power. Edge computing can mitigate the overheads for their operation by data preprocessing and smarter control, but it costs extra energy for data processing. In this talk, I will present several approaches to overcome these challenges. Switched-capacitor DC-DC converters have enabled efficient on-chip energy harvesting and power conversion in a wide power range. Digitizing internal operation of sensor interfaces and an ADC has resulted in tremendous size reductions while maintaining or even improving the energy efficiency. I will also introduce A new, efficient hardware architecture for AI processing on edge devices. In the future, increased data collection from IoT devices will burden the system with more data processing and transfer requirement. This trend will lead to integration of IoT devices, hubs, and base stations, and remote servers in a big cloud platform, where datasets collected from the leaf IoT devices are combined and refined step by step as they flow towards higher nodes. I will outline some possible research directions for further improving each IoT component and realizing a big cloud with smarter IoT systems.