Research

In our laboratory, we are researching new-paradigm hardware architectures to enable next-generation VLSI computing. Specifically, we are exploring multi-value current-mode integrated circuits, logic-in-memory VLSI architectures, and high-speed asynchronous data transfer. Multi-value current-mode integrated circuits attempt to solve the problem of wiring increasingly dense and complex circuits by enabling analog computation and transmission. Logic-in-memory VLSI computing achieves huge memory bandwidth by dispersing storage functions into arithmetic circuits. High-speed Asynchronous data transfer can perform adaptive operations that are capable of crossing clock domains. Using these core technologies, we are promoting research towards the realization of superchips for communication technology and high-performance VLSI processors for multimedia applications.

Nonvolatile Logic LSI

In order to realize advanced Internet of Things (IoT) society, it is required that edge terminals (local devices) once again perform intensive computations, while now maintaining ultra-low power consumption. However, conventional volatile circuits must be powered to store information, which consumes a large amount of standby power. Non-volatile logic solves these problems and makes it possible to completely cut off standby power during while idle, thus enabling ultra-low power consumption. In our laboratory, we are researching various nonvolatile logic circuits such as MCU and FPGA.

in-Vehicle Network Security Technology by AI

With the advancement of automotive computer systems, the security of in-vehicle networks which are the basis of driving control, remains fragile. In this research, we are researching in-vehicle network security technology that can detect various external attacks with high accuracy by using AI advanced recognition processing.

AI Processing Based on a New Computational Principle

When we recognize an object (such as a dog or a cat), it is said that the image from our eyes is broken down into components and the object is recognized as a combination of the component information. In this demo, the function of the first visual area (V1) that extracts components of visual processing in the brain is reproduced on LSI.