Research

We position ‘embedding edge AI into robots’ as our strategic core.
Our aim is to implement AI on digital circuits such as FPGAs and ASICs, and integrate it into home service robots, dual-arm robots, and any other human support robots.

01

Brain-Inspired AI

Realize a brain-inspired artificial intelligence model designed based on brain functions such as the hippocampus, amygdala, and prefrontal cortex. Develop an artificial intelligence model that acquires knowledge unique to each home environment through experience—knowledge that cannot be obtained from large-scale datasets. Integrate this model into a home service robot and implement it in dedicated integrated circuits to achieve a highly efficient system with low power consumption.
02

Digital Circuit

We aim to establish high-performance, low-power digital circuit design technologies, developing hardware accelerators and specialized processors for the AI and IoT era. Our research spans from FPGA implementation to ASIC design, creating foundational technologies for next-generation digital systems that enable edge computing and real-time processing applications.
03

Real World Robotics

We aim to apply robots to real-world environments characterized by high uncertainty. Our research focuses on achieving human-robot interaction through spatial and task understanding. We target a broad range of research areas, from general-purpose models like imitation learning to edge AI designed for immediate adaptation to specific environments, striving to solve challenges from multiple perspectives.
04

Resevoir Computing(Co-Work with Tanaka Lab.)

To establish a theory of embedded-oriented AI, this group focuses on reservoir computing (RC), a lightweight AI, and explores its optimal topology. By utilizing RC, we aim to reduce the weight of functions in current AI and realize the ability to adapt to individual environments, a weakness of current AI. We can then apply this as a fundamental technology for service robots and other applications.
05

Dataset Generation

We aim to create intelligence optimized for “the thing” through data-centric AI development, thereby enabling robots to autonomously adapt to their environments. Through data-generation methods, we advance backend technologies that make AI models more accurate, efficient, and lightweight.
06

Analog Circuit

We aim to develop edge AI chips capable of “thinking” as flexibly as humans.
We design integrated circuits that combine time-domain analog techniques with emerging non-volatile memories, and develop brain-inspired AI models based on neural dynamics, designed for hardware implementation.
07

Robot System

To accelerate the social implementation of autonomous robots, we are identifying challenges faced during the introduction and utilization process and exploring solutions. Specifically, to efficiently and smoothly deploy multifunctional and intelligent robots into society, we are promoting intelligent processing systems that prioritize minimizing introduction costs and time.