Semi-autonomous Risk Assessment System with Incidents Reporting Function

这项研究的目的是自动化一系列的风险评估过程，从风险感知到向工人提供警报信息，以确保不会遇到任何危险情况。对这项研究的讨论是基于危险点信息是已知的这一假设进行的。我们为半自动的风险评估过程开发了各种组件技术。首先，我们使用加速度计和陀螺仪构建了一个人体部位运动轨迹检测器，保证了对环境噪声的高度鲁棒性。我们还开发了一种预测器，通过获取当前位置的感觉信息和人类施加的操纵力，在几秒钟内产生输入。它们被结合到技能辅助的动力学和用于阻抗控制的状态方程中。在文本挖掘技术方面，我们实现了一个编码程序，将词组编码成其类别代码，以增强t…更多的是文本聚类或信息检索操作的性能。还开发了传感器辅助的多式联运事故报告系统(MIRS)，使人们能够直观地报告危险情况。基于m-Shel模型，所有来自传感器的记录都有助于在高度系统化的数据结构中表达事件及其原因。最后，我们对所开发的技能辅助系统的收敛进行了实验研究，以确保在提高系统安全级别后不会引发任何事故/事件：异种复制所有传感器信号通道和控制器以实现自诊断。我们根据受试者的口头信息，实验调整了预测时间，相当于危险点附近体积的扩张/收缩。最后得出结论，半径22 mm附近区域的实验结果表明，预测时间不收敛到S调整时间的0。要求在危险点周围以较小的体积进行进一步的实验。较少

The study aimed at automating a series of risk assessment process from risk perception to supplying alarming information to a worker maneuvering Skill-Assist in order not to encounter any hazardous situation. Discussion on the study proceeds based on the hypothesis that hazard point information is known at hand.We developed various component techniques toward the semi-autonomous risk assessment process. First, we built a human body part motion trajectory detector by use of an accelerometer and a gyroscope promising high robustness against environmental noise. We also developed a predictor which generated inputs in several seconds ahead through acquisition of the sensory information of the current position and the maneuvering force that human exerted. They were incorporated into the dynamics of Skill-Assist and the state equation for the impedance control. Regarding the text-mining technology, we implemented a CODING program which encodes word groups into its class codes for enhancing t … More he performance of text-clustering or information-retrieval operation. A sensor-aided Multimodal Incident Report System (MIRS) was also developed, which enabled persons to intuitively report dangerous situations. All records from sensors contribute to expressing incidents and their causal factors in a highly systematized data structure based on the m-SHEL model.Finally, we conducted experiments investigating the convergence of the developed Skill-Assist system so as not to cause any accident/incident after enhancing the safety level of the system : duplicating all sensor signal channels as well as the controllers heterogeneously for self-diagnosing purposes. We experimentally adjusted the prediction time, which was equivalent to expansion/contraction of the volume in the vicinity of hazard points, by the verbal information from the subject. It is eventually concluded that the results of the experiment under the vicinity area of 22 mm radius showed prediction time did not converge up to 0 s of adjustment time. It is demanded that further experiments be conducted with smaller volume around the hazard points. Less