Development of a Multi-Fingered Soft Gripper Digital Twin for Machine Learning-based Underactuated Control

TL;DR

The paper tackles the challenge of underactuated control for multi‑finger soft grippers by building a referencable digital twin that captures salient soft‑material phenomena, including nonlinearity, hysteresis, uncertainty, and time‑varying effects. Uncertainty is modeled with Monte Carlo sampling, and a Q‑learning framework is used to identify an actuation speed that minimizes process uncertainty and enables coordinated underactuated motion in simulation. The approach is demonstrated in a two‑finger setup, showing that high‑speed coordination yields tighter finger synchronization than slow speeds, and outlining a path toward sim‑to‑real transfer. This Digital Twin sets the stage for training advanced ML-based control strategies for soft grippers and for extending to contact mechanics and grasp‑success prediction.

Abstract

Soft robots, made from compliant materials, exhibit complex dynamics due to their flexibility and high degrees of freedom. Controlling soft robots presents significant challenges, particularly underactuation, where the number of inputs is fewer than the degrees of freedom. This research aims to develop a digital twin for multi-fingered soft grippers to advance the development of underactuation algorithms. The digital twin is designed to capture key effects observed in soft robots, such as nonlinearity, hysteresis, uncertainty, and time-varying phenomena, ensuring it closely replicates the behavior of a real-world soft gripper. Uncertainty is simulated using the Monte Carlo method. With the digital twin, a Q-learning algorithm is preliminarily applied to identify the optimal motion speed that minimizes uncertainty caused by the soft robots. Underactuated motions are successfully simulated within this environment. This digital twin paves the way for advanced machine learning algorithm training.

Figures (5)

• Figure 1: Q-learning is applied to determine the optimal motion speed for the digital twin of the soft gripper in order to achieve underactuated motions.
• Figure 2: The stress-strain curve of soft materials demonstrate the nonlinearity of several commonly used materials in soft robots. Marechal2021material.
• Figure 3: The hysteresis effect differs the forward ($\zeta = 0.7$) and backward ($\zeta = 0.8$) responses of a soft actuator in the simulator.
• Figure 4: (a) Monte Carlo method is applied to simulate the dynamics of a soft pneumatic actuator under uncertainty. (b) High speed and low speed responses of soft pneumatic actuator and their excited uncertainty. The standard deviation of the steady-state errors for low-speed motions is around 0.1 $rad$ while that of high-speed motions is around 0.05 $rad$.
• Figure 5: (a) Simulation of underactuation in the two-fingere soft gripper at high speed. (b) Simulation of underactuation in the two-fingered soft gripper at low speed.