| 則兼 卓人 | 1551076: M, 2回目発表 | ソフトウェア工学 | 松本 健一,飯田 元,Graham Neubig,伊原 彰紀 |
|
title: Early Identification of a Long-term Contributor based on Patch Submission History in Open Source Projects
abstract: In Open Source Software (OSS) projects, long-term contributors, who join for more than one year?, is a key person to maintain the software continually for a long time. To keep highly reliable software, OSS projects should secure the long-term contributors. However, some or the contributors often leave the projects, because they are a volunteering developer. The goal of my study is to identify long-term contributors based on their patch submission that is an main contribution in OSS projects. language of the presentation: Japanese | |||
| 則行 祐作 | 1551077: M, 2回目発表 | ソフトウェア工学 | 松本 健一,飯田 元,畑 秀明 |
|
title:
A Study of the Growth of Programmers with Online Judge Archives
abstract: An online judge is an online system that provides programming problems and an environment for compiling and testing submitted source code. Such systems store the histories of executions and users can see them. Programmers can learn and improve their programming skills by addressing problems in an online judge system. However, improving their skills is not easy and programmers sometimes stop learning. To help programmers learning in an online judge, we aim to understand the patterns of learning process. In this study, we analyze the histories of AOJ online judge by finding the characteristics of problems and investigating the histories of programmers' activities. language of the presentation: Japanese | |||
| 平尾 俊貴 | 1551085: M, 2回目発表 | ソフトウェア工学 | 松本 健一,飯田 元,伊原 彰紀 |
|
title: Understanding Code Review in Contentious Patches -A Case Study of the OpenStack Community- abstract:Code review is a broadly adopted software quality practice, where team members critique each others’ patches (i.e., changes to a software system). In addition to providing constructive feedback, reviewers are asked to provide a score to indicate whether the patch should be integrated into project repositories and their level of confidence. Reviewers may disagree with each other, yielding contentious patches, i.e., patches that receive both positive and negative scores. In this research, I set out to better understand contentious patches. To do so, I perform a case study of the code review process of OpenStack―a large and thriving open source community. | |||
| RADEVSKI STEVCHE | 1551129: M, 2回目発表 | ソフトウェア工学 | 松本 健一,飯田 元,畑 秀明 |
|
title: Automatic Classification and Evaluation of Code Examples
abstract: Code examples are one of the most commonly used knowledge sources when learning the usage and best practices of a new programming library. As examples are written by different developers, the provided examples will vary in style, quality, and consistency. This lack of standardization and guarantee for quality may result in demanding or improper use of a library. To resolve the problems mentioned in the previous paragraph, and to support the creation process of examples, our work focuses on the creation of an examples classification and evaluation tool. The aim of the tool is to ensure that examples follow the best-practices, style, and usage patterns of a particular library and programming language. This research has been divided into three stages: Examination of the current state of existing examples; Clustering on example features and cluster evaluation; Building an evaluation tool based on the outcome of the cluster evaluation. language of the presentation: English | |||
| 石原 弘二 | 1561003: D, 中間発表 | 知能システム制御 | 杉本 謙二,小笠原 司,森本 淳,松原 崇充,南 裕樹 |
|
title: Humanoid Robot Control with Hierarchical Model Predictive Control based on Human Skill Abstraction via Inverse Optimal Control abstract: Motion generations for various tasks under changing objectives are required for a humanoid robot. In such a case, it might be a fruitful approach to generate skillful robot behaviors with captured human movements executing the tasks since humans and humanoid robots share similar body structure. If human movement skills are extracted as objective functions, a real-time optimal control approach, Model Predictive Control (MPC) can achieve the adaptive motion generations because MPC is able to successively generate the motions satisfying the objectives. However, since the dynamics of humans and biped robots are not exactly the same, we cannot directly copy human joint angle trajectories to transfer human movement skills to humanoid robots. Moreover, real-time MPC for a robot system has been considered as impractical since MPC is computationally intensive. To cope with these problems, we, first, propose a MPC method which has a hierarchical optimization procedure to reduce the computation time of optimization. We, then, propose a control framework based on human skill abstraction: after extracting human movement skills as objective functions delineated in the abstracted state space by Inverse Optimal Control, the extracted objective functions are utilized to generate humanoid movements in the upper layer of the hierarchical MPC. language of the presentation: Japanese | |||
| 趙 崇貴 | 1551062: M, 2回目発表 | ロボティクス | 小笠原 司,池田 和司,高松 淳,丁 明 |
|
title: Hand motion recognition based on forearm deformation measured with a distance sensor array
abstract: Studies of Hand motion recognition using surface electromyogram (sEMG) signals measured from the forearm play an important role in various applications, such as human interfaces for controlling robotic exoskeletons, prosthetic hands, and evaluation of body functions. Although the sEMG signals carry a lot of information about the muscle activity, the information about deep layer muscle is limited. I focused on forearm deformation since hand motions cause not only surface and deep layer muscle deformation, but also affect the tendon and bone orientation in the forearm. Furthermore, it is likely that information about the activities of deep layer muscles can be extracted from the forearm deformation. In this study, I propose a hand motion recognition method based on the forearm deformation measured with a distance sensor array. The method uses the support vector machine. Our method achieved a mean accuracy of 95.2% for seven hand motions. Because the accuracy of the pronation and the supination are high, the distance sensor array may be able to estimate the activities of deep layer muscles. language of the presentation: Japanese | |||
| VON DRIGALSKI FELIX WOLF HANS ERICH | 1561035: D, 中間発表 | ロボティクス | 小笠原 司,杉本 謙二,高松 淳,丁 明 |
|
title: Two-armed robotic manipulation of flat deformable objects
abstract: While robots currently manipulate rigid objects with superhuman precision and strength, their performance on highly deformable objects like textiles is significantly below the public's expectations. Many tasks are either extraordinarily slow or remain unsolved problems. We identify as main difficulties the large search space for a deformed textile, as well as insufficient capabilities in most industrial grippers. This research is separated into two stages. First, we present a complete setup to fold clothes automatically, which by reducing the search space and the use of a tool for humans, achieves a reduction of the task time by 40% compared to previous work. Second, we present a new hand prototype designed for the manipulation of large textiles, specifically tucking and stretching, which are necessary manipulations for a number of unsolved automation tasks. language of the presentation: English | |||