Optoelectronic RRAM for In-Sensor Computing

Like Comment
Read the paper

      Since the publication of our paper (Nature Nanotechnology, vol.14, 776-782, July 2019, https://www.nature.com/articles/s41565-019-0501-3) last year, the concept and design of sensory computing device in our work have drawn great interest from different research fields, especially smart sensors, memory, machine vision, neuromorphic computing, etc. The in-sensor computing device, converging sensing, memory, and efficient processing functions, is an inevitable development trend in the future AI-based electronics to catch up with the demands of AI algorithms.

     The publication brings a key turning point in my career. In view of our achievements, I got a lot of chances to get the faculty positions in the university. I also, fortunately, won the 2019 Hong Kong Young Scientist Award (top two) this year, for which the final selection will be held next month. Followed with this paper, we extended the research of in-sensor computing devices for visual processing, as well as other sensory processing (e.g. olfactory, auditory, tactile, etc.). My advisor, Prof. Yang Chai, and I also summarize our thoughts as a perspective paper. The paper was then published this month on Nature Electronics (Nature Electronics, vol.3, 664-671,2020, https://www.nature.com/articles/s41928-020-00501-9). Based on the in-sensor computing device in our Nature nano paper, we proposed the near-senor and in-sensor computing paradigms in which computation tasks are moved partly to the sensory terminals. We provide possible strategies and solutions for hardware implementation of integrated sensing and computing units using novel device design and advanced manufacturing technologies. We hope that this paper can provide the readers with some ideas or directions of computing in/near the sensor with simplified circuitry, lower latency, and higher power efficiency.

      Currently, we are still working to apply our device to system/chip level applications. For example, the fabrication processes are explored for large-scale application; the designs are under optimization for temperature stable and moisture stable application; the integration with peripheral circuits are required to be performed for all hardware-implemented processing. We are also aiming to develop the in-sensor computing device for more complex and high-level processing functions. We hope that our work and views in this research field could give some directions to the community. In addition, we hope that the community can also help to improve the device design or realize the chip level application in the future.

     

Feichi Zhou

Postdoctoral Researcher, The Hong Kong Polytechnic University

No comments yet.