A Hybrid Controller for a Soft Pneumatic Manipulator Based on Model Predictive Control and Iterative Learning Control

Area of Science:

• Robotics
• Control Systems
• Soft Materials

Background:

• Soft robots offer high flexibility and dexterity, but their dynamic modeling and control are challenging.
• Existing model-based and model-free methods struggle to balance complexity and accuracy.
• Accurate dynamic models are crucial for precise control of soft robotic systems.

Purpose of the Study:

• To develop a dynamic model for a three-chamber continuous pneumatic manipulator using the modal method.
• To propose a hybrid controller integrating model predictive control (MPC) and iterative learning control (ILC) for soft robots.
• To enable simultaneous model parameter learning and trajectory tracking control.

Main Methods:

• Established a dynamic model of a three-chamber continuous pneumatic manipulator via the modal method.
• Developed a hybrid controller combining model predictive control (MPC) and iterative learning control (ILC).
• Implemented real-time model parameter learning and trajectory tracking control.

Main Results:

• The proposed hybrid controller optimized dynamic model parameters in real-time.
• Achieved superior trajectory tracking performance compared to traditional model-free methods.
• Demonstrated reduced iteration counts for model parameter learning.

Conclusions:

• The hybrid MPC-ILC controller effectively addresses challenges in soft robot dynamic modeling and control.
• The method offers a promising balance between control accuracy and computational complexity.
• Future work should validate the dynamic model and controller on multi-section manipulators.