Nurturing talents and professionals for the digital age
Results of the FY2023 MITOU IT Program
Release Date:Sep 26, 2024
In-page Links
Results of the FY2023 MITOU IT Program
PM: IGARASHI Yuki
Training Data Analysis System for Soccer Goalkeepers
This project focuses on the unique position of a soccer goalkeeper, a role that requires specialized knowledge and experience, which has led to a shortage of dedicated coaches for many teams. To address this, we developed the 'Smart Goal' system, which allows soccer goalkeepers to review their own plays. By simply attaching two cameras to the goal, the system enables 3D tracking of both the goalkeeper and field players. This significantly improves the quality of practice reviews for players of all levels, from amateur to professional.
Creators: O Hosei, MATSUO Yugo, SOMEYA Taiga
High-Accuracy Model-Based Flight Controller: Leveraging URDF for Enhanced Physical Modeling
In recent years, advancements in flight controllers, the "brains" of drones, have fueled rapid growth in the industrial drone market. Open-source technology has enabled even those without specialized expertise to develop drones. However, traditional flight controllers have limitations in supporting various drone configurations. Our project addresses this challenge by developing "Tobas," a new flight controller designed to ensure stable flight for drones with unconventional configurations that traditional products struggle to manage. Tobas comprises flight code, Gazebo plugins, and setup assistant, all designed to be user-friendly, even for those without a background in dynamics or control engineering.
Creator: DOHI Msayoshi
Kyogi Karuta for all
In this project, we have developed a Kyogi Karuta system in a virtual reality (VR) space that can be enjoyed together regardless of skill level. By using VR, it is possible to change visual and auditory information for each player, allowing for skill adjustment. In addition, the existence of the adjustments becomes less perceptible to the player and their opponent, making it harder for both players to focus on the adjustments. Furthermore, using VR has also solved the problem of providing support that would be difficult to provide in real space. This system allows users to enjoy the full range of experiences related to Kyogi Karuta in a VR space.
Creator: MARUYAMA Ayaka
PM: INAMI Masahiko
Development of a sensory map for recording and sharing information from the five senses
In this project, we developed a sensory map using a map API that allows users to post and share sensory information about specific locations to support people who face challenges in going out, moving around, shopping, or enjoying leisure activities due to sensory hypersensitivity. The sensory map enables users to check sensory information about their destination in advance on a smartphone, helping those with sensory hypersensitivity to prepare, take countermeasures, and plan escape routes.
Sensory hypersensitivity is a condition where various senses, such as sight, hearing, smell, taste, and touch, become overly sensitive, making daily life difficult. It can manifest as a symptom of various disorders and conditions, including developmental disorders, mental illnesses, autonomic nervous system disorders, and dementia. However, sensory hypersensitivity itself is not a medical diagnosis, and there are no established standard diagnostic or treatment methods. People with sensory hypersensitivity often have difficulty going out due to stimuli like light, noise, or smells, and in the case of children, this can lead to meltdowns (panic states) and significant physical exhaustion.
Traditionally, people have used measures such as earmuffs or sunglasses to cope with these challenges, but these are not complete solutions. This system is expected to provide more opportunities for outings that have been unconsciously avoided, thereby promoting social participation and improving the quality of life.
Creators: KATO Jiei, ANDO Yosuke, TERASAKI Yuki
A high-speed GUI pointing system that uses only your eyes and keyboard
Pointing devices such as mice are mainly used to operate PCs through graphical user interfaces (GUIs), but GUI operations require time to recognize the pointing target, move the cursor, and make a selection. This problem becomes more pronounced as the number of monitors and resolutions increase. On the other hand, there are tools such as Launchers and Vim that allow you to use keyboard input to control the GUI.
Using these can sometimes be more efficient than using pointing devices. However, complex GUI operations such as drag & drop cannot be performed efficiently using only keyboard input.
Therefore, this project focuses on multimodal interaction that combines the user's gaze. By estimating the user's gaze point and using it for pointing operations, we aimed to improve the efficiency of GUI operations. On the other hand, although the use of gaze estimation technology enables users to perform fast pointing without using their hands, the cost of implementation and the limitations of accuracy have been barriers to its widespread adoption.
To solve this problem, this project has developed the following two systems.
The first is a system that uses a smartphone as an eye tracker to achieve eye tracking without the need for dedicated hardware. The second is a system that combines the user's gaze and keystrokes to perform GUI operations. This system was designed to improve the efficiency of everyday operations such as selecting and focusing on elements and windows, and drag and drop. We also aimed to propose an interaction that solves the problems of accuracy and difficulty of use that currently exist in eye-tracking interfaces.
Creator: IZUMI Kazuya
Development of a Cloud-Based Reverse Lexicon from Japanese Sign Language to Japanese Utilizing Sign Language Recognition Technology
Through his activities in the sign language club, Mr. Takahashi became aware of the current situation in which beginners of sign language are having difficulty communicating with the deaf. One of the major barriers that beginners face is the difficulty of looking up the meanings of the signs used on their own. Existing sign language dictionaries and crowd services do not have sufficient reverse lookup functions for looking up the meanings of Japanese from sign language. There are two main issues in achieving this reverse lookup function. Firstly, a large data set is required to predict the meaning of words from sign language videos, and there is a lack of publicly available data sets for Japanese Sign Language. Secondly, there are significant regional and individual differences in sign language, and many non-standard forms of movement are used, but it is not realistic to cover all of these.
Therefore, this project aims to develop a cloud-based sign language dictionary that allows reverse lookup searches, incorporating posture estimation technology and findings from sign language linguistics. This dictionary will be integrated with an algorithm that utilizes collective intelligence, so that it can perform reverse lookups of meanings from various sign language movements. Special attention will be paid to the accuracy of the output, so that the reliability of the sign language dictionary is maintained and users can learn correct content. In this way, we aim to address the challenges faced by sign language learners and promote smooth communication with the deaf.
Creator: TAKAHASHI Ryota
Development of a built-in AI module for stuffed toys
This project aims to develop a dedicated embedded AI module for stuffed toys, and to apply robotics to all kinds of stuffed toys. The goal is to create a platform for efficiently embedding intelligence into a wide range of stuffed toys.
In the modern world, owning a pet can be a very big decision. In the past and now, pets still need care, but there are some unique modern circumstances, especially with dogs and cats. One is that urbanization is making it difficult to keep pets. Another is that pets are living longer. With the development of medicine, the lifespan of dogs and cats has increased dramatically in recent decades. While this is a cause for celebration, it also means that the decision to bring a pet into your life is a decision that will affect your life for the next 15 years.
Perhaps in light of this, in recent years, household dog-shaped robots have started to be sold. However, these are highly functional but extremely expensive, perhaps because they are designed with the idea of dogs as pets in mind. Therefore, in this project, we decided to develop a pet robot with the aim of embedding intelligence into a stuffed toy that does not normally move, in order to minimize any sense of incongruity even with reduced functionality.
However, even so, problems specific to stuffed toys arise. Stuffed toys are ultimately a mirror of oneself. In other words, they are objects that can project one's self, and each person has their own preferences in terms of appearance. And the fact is that there are all kinds of stuffed animals in the world, such as dogs, cats, bears, rabbits, and mice. If we tried to apply robotics to all of these, the development costs would be enormous and impractical, and even if we implemented it for some of them, it would not be enough to meet the demand.
Therefore, in this project, we will develop an embedded AI module specialized for turning stuffed animals into robots, and aim to standardize the stuffed animal skeleton. By separating the development process of the stuffed animal's fur and skeleton, we will build a foundation for efficiently embedding intelligence into a wide range of stuffed toys.
Creator: KOYAMA Takashi
PM: OKA Mizuki
Development of a smartphone application for crime prevention using GPS and synthetic voice
In this project, we have developed the iOS application “Anshin Yomichi” (Safe Night Path), which provides a sense of security to women who feel uneasy walking alone at night, and promotes cooperation with family members and partners. This application provides a new solution for keeping women safe, with features such as GPS location information acquisition and sharing, real-time communication between family members, and an emergency reporting function. Through this project, we also aim to increase social awareness of the problem of anxiety experienced by women walking alone at night.
Creators: ISHITOYA Yuri, KINJO Takuto, EBATA Keita
An underwear device for measuring menstrual blood volume aimed at alleviating stress during menstruation
In this project, we have developed “Unleak Shorts”, a panty-type menstrual blood flow measurement device, with the aim of reducing the stress of women during their menstrual periods. This device measures menstrual blood flow using sensors attached to the panties, and allows you to monitor the state of your menstrual blood flow using a dedicated smartphone application. This means that you no longer have to change your sanitary napkin as often when you are out and about, and it can reduce your anxiety about menstrual blood leakage. By freeing women from the stress of having to go to the toilet to check and the stress of constantly worrying about the state of their menstrual blood, it is hoped that this will contribute to improving the quality of life of women during their menstrual period.
Creators: YOSHIDA Saya, TSUNETSUGU Mai, TAKAYASU Yuta, NOZAKI Naoki
An application to support the creation of a plan to maximize users' satisfaction at an amusement park
In this project, we developed the “TDL/TDS AI Navi” application to support the creation of plans to maximize visitor satisfaction at Tokyo Disney Resort.
This application suggests the order for visiting attractions based on real-time crowding information and visitor preference within the park. It also uses genetic algorithms to create plans that take into account the satisfaction of all visitors. The aim of this is to make the most of the limited time spent in the park and provide an efficient and satisfying experience.
Creators: TAKEMI Kazuki, KOJIMA Sota, TONE Yukihiko
PM: OCHIAI Yoichi
Development of an Interactive Selective Weeding Robot
In this project, we worked on developing the 'OHANA-ROBOT,' a robot equipped with an interactive interface using a large language model (LLM) that autonomously performs weeding based on environmental sensing. Our goal was to create a system with high adaptability and flexibility that sets it apart from traditional agricultural robots, constantly learning and evolving through interactions with farmers. Additionally, in the hardware design of the robot itself, we emphasized the importance of enabling users to perform maintenance and customization themselves, based on the principles of open source.
Creator: KOYASU Ryuji
Development of an Espresso Extraction Control System Aimed at Maximizing Emulsification
In this project, we worked on the development of “Connoisseur”, a software/hardware that enables the control of espresso machines and the objective evaluation of extraction quality. We aimed to model the espresso extraction process as a digital twin, and through real-time sensing and simulation, to explore and reproduce the optimal extraction conditions. By doing so, we aimed to create an environment where anyone can provide consistent high-quality espresso, by visualizing the “art of the master” that had previously relied on the intuition and experience of skilled baristas.
Creator: OKADA Takuma
PM: SOGAWA Keisuke
Unikernel designed for Wasm and Wasm Compiler
WebAssembly (Wasm) is a virtual instruction set that can run on web browsers. Wasm is originally born for faster execution in web browsers than Javascript. However, using Wasm runtimes, you can run Wasm not only on browsers but also on operating systems. Unikernels are OS kernels in which a single application is statically linked with a kernel and bundled into a kernel image. Unikernels have advantages, such as lightweight, high execution speed, security, and boot time compared to ordinary operating systems because it runs only one application. This project developed a unikernel "Mewz" that would replace existing container technology by exploiting the advantages of Wasm and unikernels. We also implemented "Wasker," an AoT compiler that converts a Wasm binary into an object file, and software that can run the unikernel as if it is a container. This unikernel achieved better performance on virtual machines than existing Wasm runtimes.
Creators: UEDA Soichiro, NOZAKI Ai
Circular platform actualized by a design system using generative AI
In this project, we have developed a remake design system that takes into account the possibility of sewing and the degree of freedom in design. System users can generate clothing designs by selecting clothing parts from the system and parameters for the clothing they want. Furthermore, by implementing a 3D simulation function, we have devised a way for people who don't usually have the opportunity to see remade designs to understand what the finished design will look like.
Creators: KATO Yu, MORIGUCHI Ryota, KITSUKI Aoi
PM: TAKESAKO Yoshinori
Development of a statically typed programming language that can be transpiled to Python
In this project, we developed the statically typed programming language Erg, which can be transpiled to Python, and a set of development tools for it. The language specification is available on the official website and GitHub repository, and most of the features have already been implemented. The compiler is implemented in Rust and released as open-source software under a dual MIT and Apache 2.0 license. It is available for Linux, macOS, and Windows. As part of the development tools, we developed a language server, package manager, installer, package registry, and website that allows users to check the registered packages in the registry.
Creator: SHIBAYAMA Shunsuke
Developing electronic musical instruments that harmonize with the environment
A new musical instrument called “phonoma” has been developed based on the concept of an electronic musical instrument that plays with the environment. This instrument is a product that specializes in improvised performances in a given environment and moment in time, by capturing the state of the environment as real-time data from a LiDAR sensor and converting and outputting it as musical elements such as tone and melody.
Creators: HASEGAWA Taito
Development of a Secure and High-Performance Database System with TEE
Even in situations where cloud providers cannot be trusted, we have developed a data infrastructure called CASSA (Cloud-Adapted Secure Silo Architecture) that achieves both confidentiality and performance. This system uses Intel SGX's Scalable-SGX to alleviate the memory limitations of Enclaves, allowing them to handle large volumes of data. In addition, they have adopted Silo for the transaction processing protocol and Masstree for indexing, which has improved data processing performance and confidentiality.
Creator: FUKUYAMA Masahide
PM: TANAKA Kunihiro
NextMicon: A User-Friendly Development Environment for Custom Microcontroller Design
This project focuses on developing a user-friendly platform called ""NextMicon"" to simplify the creation of custom microcontrollers using FPGAs.
A microcontroller is a device that integrates a CPU and various peripheral circuits into a single package. For example, Arduino is a popular microcontroller development board that is widely used for many electronic projects. However, when using commercially available microcontrollers, you need to find one that meets the specific requirements of your project. By creating your own microcontroller, you can design one that is ideal for your project. This is our motivation for designing custom microcontrollers using FPGAs.
FPGAs are devices that enable users to configure custom logic circuits using programming. You can create virtually any logic circuits by programming an FPGA. However, traditional FPGA development tools can be difficult to use, especially for beginners. This project seeks to provide a user-friendly development experiment that allows people who have never worked with FPGAs before to easily and intuitively design custom microcontrollers. Designing a custom microcontroller provides useful insights into computer architecture from a hardware perspective, and it is also highly beneficial for education.
Creator: OHTA Ryosuke
Development of a system that automatically generates RISC-V-based processors
The aim of this project is to reduce the difficulty of development using FPGAs and to promote their use by more developers. While FPGAs are expected to have a wide range of applications due to their flexibility, they also have high development hurdles, such as the need for in-depth knowledge of hardware. In order to address this issue, this project has developed a RISC-V-based processor core and a coprocessor that extends it, and provides a processor generation tool for using these in combination.
The processor core implements RV32I, the basic instruction set of RISC-V, enabling highly versatile processing. On the other hand, the coprocessor provides extended instructions that are not included in the processor core, and is responsible for functions that efficiently perform specific processing. This allows developers to extend the functionality of the processor according to their needs.
Furthermore, by using the processor generation tool, it is easy to combine the processor core and coprocessor and customize them to suit the developer's own needs. This tool provides support for developers to freely design the configuration of the processor and implement it on an FPGA.
Creator: NAKAGAMI Yuta
Development of a system for visualizing the spatial distribution of fish shoals and zooplankton
The aim of this project is to develop a system for predicting the spatial distribution of fish shoals and zooplankton based on ocean physical environment data, in order to enable accurate forecasts of future catches and the implementation of fishing regulations. In the past, it has been difficult to understand the spatial distribution of fish shoals and zooplankton, and this has affected the accuracy of fishing regulations.
The background to this project is that, even though the amount of fish, which are a fishery resource, is determined by zooplankton, it is not possible to accurately grasp the amount of zooplankton resources, so sufficient analysis is not possible. Remote sensing of phytoplankton is possible using satellites and short-wave radar, but for zooplankton, it is necessary to go to the site to investigate using a fish finder or similar equipment, and it is difficult to make inferences because of the diurnal vertical migration behavior, where they stay in deep water layers during the day and move to shallow water layers at night. In addition, the species and abundance of fish must be determined by actually catching the fish, making it difficult to accurately grasp the situation in real time.
In this project, we aim to develop a spatial distribution prediction model for zooplankton and a spatial distribution prediction model for Alaska pollock based on oceanographic environmental data and spatial distribution data for fish shoals and zooplankton. This is expected to provide a new approach to predicting the spatial distribution of marine life, and enable more accurate catch forecasts and the implementation of fishing regulations.
Creators: MATSUMURA Yusaku, SATO Hiromichi
Design and Implementation of a Capability-based Microkernel for Enhanced Security
This project proposes a new approach to the design of conventional operating systems (OSs), and develops a microkernel called “Caprese” that is based on capability-based security.
Capability-based security is fundamentally different from access control lists (ACLs) and permissions based on users and groups used in UNIX-based OSs, and uses tokens called capabilities to represent access rights to resources. For example, when a web server software process listens on port 80, many UNIX-like operating systems will grant excessive root privileges to the process simply because it needs to open a port less than 1024. However, with Capability-based Security, the process is only granted the minimum necessary privileges by being granted the capability to LISTEN on port 80. In other words, if the key to a hotel room is a token, then UNIX permission management is like temporarily lending a master key to a hotel guest who has lost their room key, and Capability-based Security is like revoking the old key to the room and providing a new key.
The aim of the project is to overcome the security-related issues of conventional operating systems and to improve security and efficiency. In addition to the microkernel Caprese, the project will also implement a sample OS that runs on Caprese, as well as a generic C/C++ standard library, to demonstrate the effectiveness of Capability-based Security and explore its potential for application in real systems.
The results of this project will show a new direction for future software development by strengthening security and achieving efficient authority management, and it is expected to make a significant contribution to the development of this field.
Creator: IMAMURA Shota
Change log
-
Sep 26, 2024
Released this page