SFC Open Research Forum 2016

By the development of the Information technology, data accumulated from sensors are now utilized to make effects in the real space. However, the behavior and the needs of people changes drastically, and the detection of it is still making slow progress. Our research aims to detect the context of the real space, and to present suitable interaction to reach the analyzed data.

Many professional sports players said, in sports, practicing is important. I focused on baseball practices and found some problems. First, "amateur baseball players cannot practice alone." Second, "the places to practice batting have some limits or risks." In this research, for the problems, I propose the batting practice system "AR Batting." This system contributes to amateur baseball players practicing of batting everywhere alone.

In this projects, we propose a design tool that support making connectable and movable joint parts for personalized hardware device. Nowadays, personal fabrication has been widely known, and one of main scenarios is making personalized hardware device like panorama camera device for smart phone. However, to create personalized hardware device, there 2 problems use should be considered: (1) Designing actuation structure that enable to realize desire motion (2) Designing connector bond objects. This project aims to solve these problems by developing a design system which mainly support user to designing connectable and movable joint parts to create personalized hardware device.

In this research, we will investigate the best interaction model for aerial tele-manipulation based on extending the user vision. Aerial robotics has been attempt to introduce to tele-manipulation field, due to its low cost production, but there is a problem to secure enough information. We will create a prototype system which can orchestrate two drones with a wide range of viewpoint and which feedbacks an over-mapped picture to extend human vision.

It is becoming a popular Break dance (a part of Street dance) in recent years. Many of the system for the Street dance will be implemented, but it isn't intended for complex motion Break dance that does not yet exist. Complex movement exists particularly in Break dance. In this study, I will implement a system that specializes in Break dance.

Recent OSes have improved performance of packet I/O and IP routing table lookup. As a result, L2 and L3 protocol processing becomes a bottleneck. We address this problem by proposing a new flow caching mechanism for L3 forwarding. Unlike existing flow caches such as those used by OpenFlow, we can make data structures more compact by specifically focusing on L3 forwarding.

Recently, there has been an epidemic of big data processing. Machine learning is one of the approach to utilize big data. When put into practice, there are a lot of handcrafts and pre-processes around dataset. These non-essential processes take a lot of time and require knowledges unrelated to machine learning. To address these issue, I propose MORA a dataset managing tool for machine learning which will reduce dataset pre-processing, optimizing memory consumption and provides seamless interface.