Stability And Performance Of Human-Machine Interaction In Telerobotics

提出了一种新的主从遥控机器人控制体系结构。远程机器人中常用的体系结构是“位置误差体系结构”和“向前流体系结构”。所建议的体系结构扩展了这两个体系结构。它利用控制从机器人的主机器人和与主机器人有物理接触的人的手臂之间的能量和信息信号交换。人类是机器人的主人，因此不可避免地要进行动力传递，并由人类发出信息信号来帮助控制机器。主机的局部力反馈可以增加主机对人类输入命令的明显灵敏度。在本案例中，力信息是在两个方向上传递的。在主人的稳定性要求和人类控制的要求之间保持平衡。主控运动部分是由于人力的传递，部分是由于控制器计算机产生的命令。控制器的设计是基于人体手臂和环境的动态模型。将人的手臂动态行为建模为一组输入和输出之间的函数关系。模型组件的内部结构和操作被隐式地解释。估计器使用学习模型创建人体手臂动态模型。它接受人体肌电图（EMG）信号、手臂方向和接触力的信息。利用实验验证的人体手臂（EMG）阻抗数据库在手臂方向区域的动态行为来得到一个时变的人体灵敏度函数。学习模型解释了未校准的区域。学习模型使用肌电信号的测量（时域或频域）作为指标来推导人类阻抗作为认知命令的函数。系统性能和系统稳定性的变化将作为人类和环境负荷动态变化的函数进行实验研究。将结果与远程呈现控制系统性能时获得的类似结果进行比较。该体系结构允许将变量方便地映射到远程机器人设计规范中。这项研究推进了一种基于知识的方法，并将导致对人类动力学和约束在遥控机器人系统控制和其他类型的人机交互中的作用的理解。

This research introduces a new control architecture for master- slave telerobotic systems. Common architectures used in telerobotics are the "position error architecture" and the "forward flow architecture." The proposed architecture extends these two architectures. It utilizes the exchange of power and information signals between a master robot, which controls a slave robot, and the human arm that is in physical contact with the master. The human wears the master robot, so power transfer is unavoidable and information signals from the human help to control the machine. Local force feedback at the master can increase the apparent sensitivity of the master to human input commands. In the present case, force information is communicated in both directions. Balance is maintained between the stability requirements of the master and the requirements for human control. Master motion is partially due to the transfer of human power and partially due to the command generated by the controller computer. The design of the controller is based on dynamic models of the human arm and the environment. The human arm dynamic behavior is modeled as a functional relationship between a set of input and outputs. The internal structures and operations of the model components are implicitly accounted for. An estimator creates the human arm dynamic model using a learning model. It accepts information using human electromyograph (EMG) signals, arm orientations and contact forces. Dynamic behavior of the database of experimentally verified human arm (EMG) impedances for regions of arm orientation is used to arrive at a time-varying, human sensitivity function. A learning model accounts for non-calibrated regions. The learning model uses a measure of the EMG signals (time or frequency domain) as a index to derive human impedance as a function of a cognitive command. Changes in system performance and system stability will be investigated experimentally as a function of changes in the human and the environmental load dynamics. The results will be compared to similar results obtained when telepresence governs the system performance. The architecture allows convenient mapping of variables into telerobotic design specifications. This research advances a knowledge-based methodology and will lead to an understanding of the role of human dynamics and constraints in the control of telerobotic systems and in other types of human-computer interactions.