基于弯曲度预测模型的软体手人机交互控制

Human-computer Interaction Control of Soft Hand Based on Bending Prediction Model

软体手作为近年来机器人领域的热点研究方向,凭借较好的环境适应性和耦合安全性获得了广泛的应用。但柔性材料驱动带来的多自由度特性同时增加了软体手的操控难度。针对此问题,该文提出了一种驱动传感一体化的仿生软体驱动器用于软体手的设计与制作,顺应性传感融入为软体手操控提供了有效的传感反馈。在此基础上,利用多项式拟合与LSTM两种学习算法完成了软体手弯曲度感知的精确建模;并基于弯曲度预测模型和数据手套构建了一套软体手人机交互系统,实现了人手张合程度和软体手抓握的同步映射控制。最后,使用该系统对形态各异的目标物进行了抓取测试。实验结果验证了该文所提出软体手的实用性及基于弯曲度模型软体交互控制方法的可行性。

As a hot research direction in the field of robotics in recent years,soft hand has been widely used due to its good environmental adaptability and coupling safety.However,the multi-degree-of-freedom characteristics brought by flexible materials increase the control difficulty of soft hand.To solve this problem,a bionic software driver with drive and sensing integration is proposed for the design and manufacture of the soft hand.The integration of compliant sensing provides effective sensor feedback for the control of the soft hand.On this basis,polynomial fitting and LSTM learning algorithms are used to complete the precise modeling of soft hand bending perception.Based on the flexural sensing model and the data glove,a software hand human-computer interaction system is constructed to realize the synchronous mapping control between the degree of hand opening and grasp of the software hand.Finally,the system is used to test the grasping of objects with different shapes.The experimental results verify the practicability of the flexible hand and the feasibility of the soft interactive control method based on the bending prediction model.