A Comparative Analysis on Joint Importance to Achieve Better Performance for Future Forecasted Human Activities and Behavior Analysis for Intimate Distance Supportive HRC Sytem

• Item type: SIG Technical Reports(1)
• 公開日: 2022-05-30
• タイトル: A Comparative Analysis on Joint Importance to Achieve Better Performance for Future Forecasted Human Activities and Behavior Analysis for Intimate Distance Supportive HRC Sytem
• タイトル / 言語: en
• タイトル: A Comparative Analysis on Joint Importance to Achieve Better Performance for Future Forecasted Human Activities and Behavior Analysis for Intimate Distance Supportive HRC Sytem
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: インタフェース・状態推定
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: Kyushu Institute of Technology
• 著者所属: Kyushu Institute of Technology
• 著者所属(英): en
• 著者所属(英): Kyushu Institute of Technology
• 著者所属(英): en
• 著者所属(英): Kyushu Institute of Technology
• 著者名: Nazmun, Nahid; Sozo, Inoue
• 著者名(英): Nazmun, Nahid; Sozo, Inoue
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Human Robot Collaboration (HRC) has always been a challenging field due to safety concerns as human motions are often unpredictable and susceptible to environmental and physio-psychological changes. Previous studies aimed at predicting human behavioral trajectories focused on predicting behaviors utilizing full-body data for HRC with minimal contact with the robot. In this work, we focused on the Hand over other body parts as it will be in very close and frequent contact with the robot for complex and challenging tasks that we considered for intimate distance supportive HRC. In our analysis, we found that joint reduction and the perfect joint combination lead to better performance and joint importance varies based on the task pattern. We also successfully forecasted the human motion and activity 1s ahead using LSTM with the highest RMSE error no more than 21mm which outperformed the work in [1].
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Human Robot Collaboration (HRC) has always been a challenging field due to safety concerns as human motions are often unpredictable and susceptible to environmental and physio-psychological changes. Previous studies aimed at predicting human behavioral trajectories focused on predicting behaviors utilizing full-body data for HRC with minimal contact with the robot. In this work, we focused on the Hand over other body parts as it will be in very close and frequent contact with the robot for complex and challenging tasks that we considered for intimate distance supportive HRC. In our analysis, we found that joint reduction and the perfect joint combination lead to better performance and joint importance varies based on the task pattern. We also successfully forecasted the human motion and activity 1s ahead using LSTM with the highest RMSE error no more than 21mm which outperformed the work in [1].
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AA11838947
• 書誌情報: 研究報告ユビキタスコンピューティングシステム（UBI）; 巻 2022-UBI-74, 号 5, p. 1-7, 発行日 2022-05-30
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8698
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会