Department of Robotics, Graduate School of Science and Engineering, Ritsumeikan University, Shiga 525-8577, Japan.
Research Organization of Science and Technology, Ritsumeikan University, Shiga 525-8577, Japan.
Sensors (Basel). 2019 Sep 19;19(18):4056. doi: 10.3390/s19184056.
Soft tactile sensors have been applied to robotic grippers for assembly. It is a challenging task to obtain contact information and object orientation using tactile sensors during grasping. Currently, the design of Hall-effect-based tactile sensors to perform such tasks is based on trial and error. We present a method of investigating the optimal geometrical design of a cylindrical soft sensor to increase its sensitivity. The finite element model of a soft fingertip was constructed in Abaqus with two design variables, i.e., hollow radius and magnet position. Then, the model was imported into Isight, with the maximisation of magnet displacement as the objective function. We found that the optimal design was at the boundary of the parameter design space. Four fingertips were fabricated with one intuitive, one optimal, and two optional sets of parameters. Experiments were performed, and object orientation was estimated by utilising linear approximation and a machine learning approach. Good agreements were achieved between optimisation and experiments. The results revealed that the estimated average error in object orientation was decreased by the optimised fingertip design. Furthermore, the 3-axis forces could successfully be estimated based on sensor outputs.
软触觉传感器已被应用于机器人夹具进行装配。在抓取过程中，使用触觉传感器获取接触信息和物体方向是一项具有挑战性的任务。目前，基于霍尔效应的触觉传感器的设计用于执行此类任务是基于反复试验的。我们提出了一种研究圆柱形软传感器最佳几何设计以提高其灵敏度的方法。使用两个设计变量，即空心半径和磁体位置，在 Abaqus 中构建了软指尖的有限元模型。然后，将模型导入到 Isight 中，以磁体位移最大化为目标函数。我们发现最优设计位于参数设计空间的边界处。用一套直观的、一套最优的和两套可选的参数制造了四个指尖。进行了实验，并利用线性逼近和机器学习方法估计了物体的方向。优化和实验之间取得了很好的一致性。结果表明，通过优化指尖设计，物体方向的估计平均误差减小了。此外，还可以根据传感器输出成功估计 3 轴力。
