※ Affiliations and titles are as of the end of the research activity.

Takuya Azumi

Software platform for many-core processors to realize for processing mass date in real-time

Researcher

Takuya Azumi

Outline

Processing with mass data such as the localization and detection for an autonomous vehicle is required to process in real-time. This research aims to construct a software platform that can efficiently develop programs for many-core processors. The proposed platform realizes high-performance processing with low power consumption.

Hiromi Arai

Privacy preserving data mining for secure and transparent personalization

Researcher

Hiromi Arai

Outline

This project aims to develop algorithms and protocols for privacy preserving, secure, and transparent personalized services. Based on machine learning and cryptographic techniques, we will develop methods to prevent adversarial data provision while preserving honest users’ privacy. We will also develop data privacy preserving data publish methods for transparent personalized services.

Naoyuki Kamiyama

Studies on discrete structures in mathematical system design and its applications

Researcher

Naoyuki Kamiyama

Outline

The aim of this research is to construct fundamental theoretical tools for mathematical system design in social issues (e.g., mechanism design in societies and urban planning), and to apply our techniques for real problems. Especially, we focus on theoretically and practically hard problems with bad structure and integration with techniques in other fields, e.g., simulation.

Mitsuhiro Kamezaki

Interactive proximal navigation for human-symbiotic mobilities

Researcher

Mitsuhiro Kamezaki

Outline

The purpose of this research is to design an interactive proximal navigation method for human-symbiotic mobilities, which can freely move through human-existing environments, in particular, crowds such as train stations and airports. In order to realize safe, secure, and efficient human-mobility coordination, the navigation system will consist of observation, cognition, and control modules for inducing mutual approximation and physical contact.

Sho Sakaino

End to End Learning for Unknown Object Manipulation Using position and Force Information

Researcher

Sho Sakaino

Outline

Manipulation of unknown objects whose shapes, rigidity, and so on are not identified in advanced, is a big problem for robotization of manual labor. In this research, position control, which is robust against disturbances, and force control, which has excellent adaptability against environments, are integrated to construct object manipulation AI. First data related to object manipulation is collected. Then, actions of robots are directly leaned from the dataset.

Edgar Simo-Serra

Artificial Intelligence for Augmented Digital Creation through Interactive Personalization

Researcher

Edgar Simo-Serra

Outline

Repetitive tasks are one of the main reasons for low productivity in content creation industries such as animation. The proposed research is centered on reducing and minimizing the amount of repetitive tasks in content creation through the use of Artificial Intelligence (AI). A focus will be done on augmenting existing creation pipelines, such as illustration or photography editing, such that the creator can remain in control of the creation while increasing productivity with semi-automatic and intelligent tools. Furthermore, the AI will learn from the creator as s/he creates new works in order to personalize the AI and increase it’s performance, and adapt to the different necessities of each creator.

Jinjia Zhou

A Highly-Distributed Surveillance System with Batteryless and Wireless Video Acquisition

Researcher

Jinjia Zhou

Outline

More and more smart homes and smart buildings will be equipped with billions of ubiquitous surveillance cameras. However, with the increasing number of wireless video cameras, the surveillance systems including video capturing, compression and transmission, are going to be challenged by the power efficiency. This research proposes a highly-distributed surveillance system with batteryless video acquisition.

Ryo Nagata

Development of Automated Comment Generation Techniques for a Novel Learning Environment

Researcher

Ryo Nagata

Outline

Learners always do learning exercises, solving quizzes and questions. However, they often receive little or even no feedback because it takes a lot of time to give them feedback. In this research project, I will explore techniques to automatically generate comments on the writings of learners of English. I will also explore techniques to retrieve information, such as textbooks, related to what learners write. These techniques will contribute to realizing a novel learning environment.

Takuro Fukunaga

Adaptive optimization for decision making from dynamic and uncertain data

Researcher

Takuro Fukunaga

Outline

The goal of this research is to develop new methods for efficient decision making from dynamic and uncertain data, including data generated by sensors and machine learning. In particular, we focus on the adaptive optimization. We aim to build a basic theory for designing and analyzing adaptive optimization algorithms as well as its applications to practical problems arising from industry.

Takuya Matsuzaki

Automatic Text Layout based on Language Technology for Assisting Students with Difficulty in Text Comprehension

Researcher

Takuya Matsuzaki

Outline

The readability of a text depends highly on its layout. The aim of this research is to develop a technology to make a text more readable by automatically adjusting its layout on electronic devices. The target is the digital textbooks, which is expected to be adopted gradually in coming years. The ultimate goal of the research is not just to assist students with difficulty in text comprehension for various reasons but to foster their reading skill through their experience of reading by themselves. The layout style is selected among various options according to each student’s reading skill. A model of reading skill development will be constructed based on large-scale experiments and it will support the choice of an appropriate layout style.