| GARCIA RICARDEZ GUSTAVO ALFONSO | 1151129: M, 2回目発表 | 小笠原司, 中村哲, 高松淳, 山口明彦 |
|
title:Reactive Strategies for Human Safety during Human-Robot Interaction: an Approach to Dependable Humanoid Robots
abstract:Robots have become more and more skillful, but still in most cases their operation remains to be limited to controlled environments. Recently, there has been an increasing tendency of looking for ways to deploy robots in daily-life environments. That is why, in order to make the Human-Robot Interaction (HRI) physically closer, a methodology to improve human safety is necessary. One way of improving human safety is to make the robot react to changes in the environment by following a strategy. Therefore, I propose a methodology to provide human safety during HRI, which consists of three reactive strategies for human safety: Emergency Stop, Velocity Moderation, and Withdrawal. language of the presentation:English | ||
| 赤塚祥太 | 1151003: M, 2回目発表 | 杉本謙二, 飯田元, 平田健太郎, 野田賢 |
|
title:Similarity Analysis of Sequential Alarms in Plant Operation Data by Levenshtein Distance
abstract: The sophistication of the plant alarm system has made it possible to install many alarms cheaply and easily. As a result, alarm system has designed without the discussion of necessity and validity, and it cause frequently plant accidents by sequential alarms. It is necessary to reject the sequential alarms that extract what kind of sequential alarms occur in plant log data. However, it is not realistic that we analysis all sequential alarms one by one because there are thousands of types alarms set in a plant. This study’s purpose is the similarity analysis of sequential alarms by “Levenshtein Distance” which is widely used in biology to measure the variation between DNA. The Levenshtein distance is a string metric for measuring the amount of difference between two sequences, which is defined as the minimum number of edits needed to transform a string into the other with edit operations such as insertion, deletion, and substitution of a single character. In this presentation, I’ll show the proposed analysis method by Levenshtein Distance and the results of similarity analysis of sequential alarms which was generated by plant simulator. language of the presentation: Japanese 発表題目: Levenshtein距離を用いたプラント運転データにおける連鎖アラームの類似性評価 発表概要:近年のプラントアラームシステムの高度化により, 低コストで手軽にアラームを設定することが可能となった. しかしその結果, 必要性や妥当性が十分に議論されないままアラームシステムが設計され, 連鎖アラームなど「アラームの洪水」によるプラント事故が頻繁に発生している. 連鎖アラームを削減するためには, プラント運転ログデータの解析によってどのような連鎖アラームが発生しているのかを抽出することが必要となる. しかしログデータ中には数千種類ものアラームが存在しており, それらを一つ一つ解析するには多大な時間と労力が必要になり, 現実的ではない. この解決策として, 発生パターンの類似性が高い連鎖アラーム同士をグルーピングすることで, 解析作業の効率化を図るという手段が考えられる. そこで本研究では, Levenshtein距離を用いて連鎖アラームどうしの類似性評価を行う. Levenshtein距離とは, DNAやゲノムの類似度解析に広く活用されている, 二つの文字列間の類似度を測る基準である. 挿入・削除・置換という3つの操作を行い, 一つ目の文字列をもう一方の文字列へと変換するコストの総量として出力されたLevenshtein距離を, 連鎖アラームの類似度評価に用いる. 本発表ではLevenshtein距離を用いた類似度評価方法と, 本手法をプラントシミュレータにより生成されたデータに適用した結果について述べる. | ||
| Gemalyn Dacillo Abrajano | 1161028: D, 中間発表 | 岡田実, 関浩之, 東野武史 |
|
title: Compressed Sensing-based Rainfall Field Reconstruction using Signal
Attenuation in Microwave Mesh Network abstract: This research aims to use the signal attenuation in microwave mesh network to determine the presence and location of rainfall, using a compressed sensing-based algorithm to reconstruct the rain field. In this research, we specifically look at the attenuation due to rain of 25 GHz microwave networks, with an additional goal of using the same networks as alternative rain sensors. Link designs incorporate the possibility of rain fades so that there will be a minimal loss of communication during a rain event, and the same attenuation data can be used as input in a compressed sensing reconstruction to determine where the rain is and in what direction it is moving. Network simulations were carried out using real rainfall data from different parts of Japan to show how links with different angular separation and orientation behave under different climate. In the tropical region, the rain field as seen by the network is sparse, one of the requirements of compressed sensing, and the reconstruction can be done using linear programming. The simulations also aim to detect the presence of intense localized rains that can trigger landslides and flash floods. Initial results show that the compressed sensing algorithm can detect extreme rain events. The study has possible applications on rain detection and forecasting, identification of the location of extreme rain events without having to use conventional weather sensors, and in designing rain-induced-disaster warning and response systems. language of the presentation: English | ||