Kiyoshi IRIE

Senior Research Scientist - Future Robotics Technology Center, Chiba Institute of Technology

Hello, I am a researcher working on various types of robots. My research interests are in the field of mobile robots, machine learning and computer vision.
My goal is to create robots that make our life easier. Recently I completed my Ph.D. program at Tokyo Institute of Technology (supervisor: Prof. Masashi Sugiyama).


CV at Researchmap


IMU-based Motion Estimation

Nowadays, compact and cheap inertial measurement units (IMUs) are widely used such as in smartphones. We aim to develop practical methods for measuring motions solely from IMU data, with the hope that they will be helpful for learning sports or other skills. In theory, motions can be estimated through integration of acceleration and angular velocity measurements from an IMU; however, we need to compensate large accumulated errors to achieve practical accuracy. We developed an algorithm to reduce accumulated errors through fusing multiple constraints, which is inspired by graph-based SLAM technique.

Journal paper:

Mobile Robot Localization Using Street Maps

Recent advances in SLAM enabled robust and accurate robot localization. However, mapping a large environment is very costly. We have been working on localization by reusing existing 2D street maps (e.g., Google Maps). When using a street map, making a correct correspondence between sensor data and the map is a challenge because street maps contain very limited information. We have proposed object-recognition-based approach (ICRA2012, IROS2013) and dependence maximization approach (IROS2015).

Journal paper:

Conference papers:

Vision-based Outdoor Localization

Vision-based outdoor localization is a challenge because image local features are greatly affected by illumination conditions. We have proposed two types of methods to overcome the issue: stereo-based approach (IROS2010) and HDR vision-based approach (ICRA2011).

Journal paper:

Conference papers:

Camera-LiDAR Extrinsic Calibration

Camera-LiDAR systems became popular for outdoor RGB+D data acquisition. We developed a method for calibrating the relative pose between a camera and a LiDAR using statistical dependence maximization. Our method performs robustly even on noisy outdoor data sets.

Conference paper:

Real World Robot Challenge (Tsukuba Challenge)

We have built outdoor mobile robots and performed navigation experiments in real urban environments. We participated in Tsukuba Challenge in 2009-2011, and we were awarded Tsukuba City Mayor Award.


I have been a member of a Robocup team CIT Brains since 2007. We are the winner of Humanoid League (KidSize) at 2014 and 2015. My contributions are software components on localization, perception, planning, system architecture, GUI, etc.

(c) 2015-2016 Kiyoshi Irie