Force Estimation and Position Control of a Hydraulic Folded Pouch Actuator for Soft Robotics
Jie Li, Jianlin Yang, Jinling Qiu, Hanqi Lou, Zhangxi Zhou, George Mylonas
AI summary
Problem
Folded pouch actuators offer large strokes but suffer from strong nonlinearities and angular hysteresis, making accurate position control and force estimation difficult without complex explicit modeling.
Approach
The authors implement a sensor-based closed-loop controller with dynamically tuned PID gains and an MLP feedforward predictor for position tracking, while training MLP and LSTM neural networks on internal pressure, volume, and angle data to estimate external force.
Key results
- Closed-loop tracking achieves 4.82° MAE and 5.48° RMSE under constant loads
- LSTM force estimator outperforms MLP with 0.96 mN·m MAE under dynamic loads
- Dynamically tuned PID gains with MLP feedforward compensate for hysteresis and stress relaxation
- Data-driven framework eliminates need for explicit nonlinear system modeling
Why it matters
Provides a practical, model-free control and sensing framework that advances the deployment of soft hydraulic actuators in wearable and surgical robotics.
Abstract
This paper investigates position control and force estimation for a hydraulic folded pouch actuator. First, exper- imental platforms are designed to characterize the actuator and the results show two key properties: (i) angular hysteresis when the motion direction reverses, and (ii) strong nonlinear- ity between liquid volume, pressure, and angle. For position control, we explore three strategies: fully open-loop control, observer-based control, and sensor-based closed-loop control with angle feedback. The closed-loop controller employs dy- namically tuned PID gains and an MLP feedforward predictor. Under a sinusoidal reference, the closed-loop controller achieves mean absolute error (MAE) = 4.82° and root mean square error (RMSE) = 5.48°. For force estimation, we train both MLP and LSTM models using liquid volume, angle, pressure, and angular rate as features to predict the external force on the actuator. Compared to the MLP, the LSTM incorporates temporal dynamics, which allows it to capture force variations more effectively and generate smoother prediction results. Under dynamic loads, both models capture the applied force, with the LSTM yielding the lower errors (MAE = 0.96 mN·m, RMSE = 1.23 mN·m).