コロキアムB発表

日時: 7月17日(火)5限(16:50~18:20)


会場: L1

司会: 黄 銘
HAN XINYOU M, 1回目発表 知能システム制御 杉本 謙二
title: Feedback Error Learning Control based on an SPR Condition: a Case Study
abstract: Feedback error learning is a powerful tool for designing 2DOF (two-degree-of-freedom) control systems with an online-tuning feedforward controller. A new tuning law based on the Lyapunov function has been proposed in the context of sensing failure tolerance. It requires, however, an SPR (strictly positive real) condition. In this research, simulation results will be analyzed and imporved by changing some parameters. Also, results of the proposed method will be compared with that of conventional method to show the advantages of the proposed method.
language of the presentation: Japanese
 
MEI WENJIE M, 1回目発表 知能システム制御 杉本 謙二
title: Instability Analysis of Markov Jump Linear Systems by Spectral Optimizatino
abstract: We are concerned with the stability property of continuous-time Markov jump linear systems. Although there exists several effective criteria for analyzing the mean stability of Markov jump linear systems, there is scarse of methodologies for verifying their instability. To fill in this gap, we present a novel condition for verifying if a given Markov jump linear system is not mean stable. In deriving the condition, we construct an auxiliary Markov jump linear system from the given system as well as appropriate weight matrices. Furthermore, we prove that our condition improves an existing instability criterion.
language of the presentation: English
 
ZHU LINGWEI M, 1回目発表 知能システム制御 杉本 謙二
title: Reinforcement Learning for Plant-wide Control
abstract: Reinforcement learning, as an important part of machine learning, has been gaining remarkable results and attentions recently. No matter in AlphaGo of Google DeepMind, or autonomous robot control, we can see its existence. This research focuses on using reinforcement learning to achieve autonomous control in large scale system like factories. We analyze the difficulties with using reinforcement learning techniques for such large scale, high dimensional problems, and then describe our approach of disposal: we propose a new reinforcement learning algorithm which can successfully handle the difficulties that traditional methods fail to do.
language of the presentation: English
 

会場: L2

司会: Raula G. Kula
MAIPRADIT ARNAN M, 1回目発表 モバイルコンピューティング 伊藤 実
Title: Traffic Signal Control Based on Back-Pressure Algorithm and LSTM (Long Short-Term Memory)
Abstract: Back-pressure algorithm has been increasingly attractive for reducing traffic congestion in road networks. Recent work has shown the efficiency of back-pressure based traffic signal control algorithms. However, these back-pressure algorithms control traffic signals based on only current traffic information , which is short-sighted. A better one should also consider future traffic information. In this research, we propose a back-pressure based traffic signal control algorithm that uses both real-time and predicted traffic information . Specifically, we use LSTM (Long and Short-term Memory) to predict traffic information , i.e., number of vehicles at one road in next time period. As validated by simulations, our algorithm predicts number of vehicles at one road with high accuracy. In the future, we will run simulations to verify the efficiency of our LSTM based back-pressure algorithm in SUMO and propose dynamic vehicle routing protocol to further reduce traffic congestion.
Language of the presentation: English
 
MAIPRADIT RUNGROJ M, 1回目発表 ソフトウェア工学 松本 健一
Title: Identifying “On-Hold” Self-Admitted Technical Debt
Abstract: Self-admitted technical debt refers to situations where a software developer knows that their current implementation is not optimal and indicates this using a source code comment. Based on a qualitative study we first identify one particular class of debt amenable to automated management: “on-hold” self-admitted technical debt, i.e., debt which contains a condition to indicate that a developer is waiting for a certain event or an updated functionality having been implemented elsewhere. We then design and evaluate an automated classifier which can automatically detect these “on-hold” instances with a precision of 0.81 as well as detect the specific conditions that developers are waiting for.
Language of the presentation: English
 
NUKULKIT PONGBHOP M, 1回目発表 ソフトウェア工学 松本 健一
Title: Understanding how documentation changes with code – a case study of GitHub Readme updates
Open Source Software (OSS) projects hosted in platforms such as GitHub often release documentation (i.e. Readme) containing important information for potential and existing users and developers. Because of developers struggle for keeping documentation consistency. In this presentation, I will discuss our investigating to understand how the readme file is updated with code overtime. We mine the history of a OSS project in GitHub with over 8,000 commits for mining information how developers update their readme file. Early results show that 82 percent of dependent changes with Readme is build file. This work is towards to build notification tool for helping develops keep their Readme file up to date.
Language of the presentation: English