軟性機器手掌開發及雙手掌手臂系統控制

Abstract

軟性材料因彈性使得軟手指的自然地順應著外力改變彎曲角度而無需而外使用複雜的控制方式來達到相同效果，其天生的被動順應性確保了機器人與人類及環境互動的安全性。故軟性材料在在多指機器手的領域愈來愈受重視。由於目前的研究比較缺乏整合軟性致動器設計、仿人手設計、手指運動模型、坐標系建立、控制架構以及氣動與通訊系統在一起，因此本論文專注於整個軟手系統的完整性。本論文設計與分析了採用3D列印技術製作的輕量、新型的多指軟性仿人機械手。本論文也提出了軟性機械手的運動模型與座標系統的建立方式來建立靜力學與動力學的模擬環境。彎曲角度對應氣壓模型的平均誤差為5.67度。指尖力對應氣壓模型的誤差則為10 ~ 50 kPa。除此之外，藉由提出的手指運動模型本論文提供了兩種基於模型的控制架構分別為有感測器回授與無感測器回授。為了讓手掌與理論架構能順利實現，通訊系統與氣動系統也在論文中呈現。最後在模擬及實驗中呈現了軟性仿人機械手的性能、控制器架構之間的表現差異、不同抓握方式的展現以及軟性仿人機械手與人類及環境互動的情境。實驗結果展示了軟性的仿人機器人手可以抓取多種物體，無需在其上附加額外的傳感器。同時人類也可以安全地與它們互動而不會受到傷害。

The elasticity of soft materials enables the naturally changing bending angle according to external forces without additional, complex control algorithms, which guarantees the safety of the interaction between robot hands and humans. Therefore, multi-fingered robot hands become more and more popular in this field. Due to the lack of integration of soft actuator and anthropomorphic design, finger motion model, coordinate system establishment, control architecture, and pneumatic and communication systems in current research, this thesis focuses on the integration of the entire soft anthropomorphic hand system. In this thesis, lightweight and novel multi-fingered, soft anthropomorphic robot hands manufactured with 3D printing are developed and analyzed. The discretized coordinate frames and model of the soft actuator motion are proposed to establish the kinematics and dynamics of robot simulation environments. The average error of the model of the bending angle corresponding to air pressure is 5.67 degrees. The error of the model of the blocking force corresponding to air pressure is 10 ~ 50 kPa. Furthermore, according to the model proposed in this thesis, sensor-based and sensorless model-based control structures can be developed. Communication and pneumatic systems are also established in the thesis. Finally, the simulations and experimental results are presented, which include descriptions of the performance of the soft anthropomorphic robot hands, the difference between the two control structures, the demonstration of different grasping types, and the applications for the soft anthropomorphic robot hands. The experiments show that soft anthropomorphic robot hands can grasp various objects without attaching additional sensors to them. Humans also can safely interact with them without harm.