Dynamic Modeling and State Estimation of Cable-Conduit Actuation During Interaction With Nonpassive Environments
Dynamic Modeling and State Estimation of Cable-Conduit Actuation During Interaction With Nonpassive Environments
复制标题
DOI：
10.1109/tmech.2020.3040159
复制
发表时间：
2021-10
期刊：
影响因子：
--
通讯作者：
Stephen Buchanan;F. Sergi
中科院分区：
文献类型：
--
作者：
Stephen Buchanan;F. Sergi
电缆导管传输等远程驱动解决方案有利于可穿戴机器人技术，有助于减少远端关节的动态负载。然而，这些系统通常会引入高反射阻抗或需要刚性传感器来进行力控制，这阻碍了它们在可穿戴设备中的集成。通过将分布式电缆应变传感与动态建模相结合来估计输出力和位置，可以实现低阻抗和易于集成。在本文中，我们提出了一种新的计算模型来分析电缆管道系统的动力学。该模型具有传动装置内运动的双向传播的特点，可以模拟人类交互系统，其中人类和机器人之一或两者都可以是力或位置源。此外，我们提出了一种基于非线性方程组迭代线性化的快速求解方程组的新方法。通过涉及与人类受试者的物理交互的实验，模型和求解方法在物理原型中得到验证。最后，我们开发了基于模型的电缆张力估计方法，给出了嵌入在传输中的多个噪声应变传感器的测量值，并量化了通过不同方法可实现的精度，作为传感器数量和位置的函数。结果表明，该模型准确地预测了原型中观察到的行为。此外，新开发的迭代线性化求解方法使计算速度比标准求解器提高了100倍。最后，我们证明，当增加传感器数量时，可以以更高的精度估计电缆张力，但如果传输的输出部分没有仪表化，则精度会降低。
Remote actuation solutions such as cable-conduit transmissions are beneficial in wearable robotics to reduce dynamic loading on distal joints. However, these systems often introduce high reflected impedance or require rigid sensors for force control that hinder their integration in wearables. Low impedance and ease of integration could be both obtained by combining distributed cable strain sensing with dynamic modeling to estimate output force and position. In this article, we present a new computational model to analyze the dynamics of cable-conduit systems. The model features bidirectional propagation of motion within the transmission, which allows for simulation of human-interacting systems, where either or both the human and the robot can be force or position sources. Moreover, we present a new method for rapidly solving for the system of equations based on iterative linearization of the system of nonlinear equations. The model and the solution method are validated in a physical prototype through experiments involving physical interaction with a human subject. Finally, we develop methods for model-based estimation of cable tension given measurements from multiple noisy strain sensors embedded in the transmission, and quantify the accuracy achievable via different methods as a function of the number and location of sensors. Results demonstrate that the model accurately predicts behavior observed in the prototype. Moreover, the newly developed iterative linearization solution method allows a 100-fold increase of computation speed compared to a standard solver. Finally, we demonstrate that cable tension can be estimated with increasing accuracy when increasing the number of sensors, but accuracy decreases if the output portion of the transmission is not instrumented.