Learning-Based Fingertip Force Estimation for Soft Wearable Hand Robot With Tendon-Sheath Mechanism

Soft wearable hand robots with tendon-sheath mechanisms are being actively developed to assist people with lost hand mobility. For these robots, accurately estimating fingertip forces leads to successful object grasping. An approach can utilize information from actuators assuming quasi-static environments. However, non-linearity and hysteresis with regards to the dynamic changes of the tendon-sheath mechanism hinder accurate fingertip force estimation. This paper proposes a learning-based method to estimate fingertip forces by integrating dynamic information of motor encoders, wire tension, and sheath bending angles. The model is modified from Long Short-Term Memory by incorporating a residual term that governs the dynamic changes in sheath bending angles. Using a tendon-driven soft wearable hand robot, the proposed model obtained RMSE less than 0.44 N. It was further evaluated under criteria ranging from different object sizes, bending angle ranges, and forces. Finally, a repeatability test (0.46 N in RMSE), real-time applicability (125 Hz), and force control (12.7% in MAPE) were performed to verify the feasibility of the proposed method.