Using backward adjustment with model predictive control for adaptive control of nonlinear soft artificial muscle

Area of Science:

• Robotics
• Control Systems
• Materials Science

Background:

• Soft artificial muscles offer compliance and safety for wearables and unstructured environments.
• Accurate modeling of soft actuator nonlinearity remains a significant challenge.

Purpose of the Study:

• To develop an adaptive control method for soft artificial muscles.
• To improve the accuracy of dynamic models for soft actuators.
• To achieve precise reference tracking in soft robotic systems.

Main Methods:

• Leveraging model learning and backward adjustment of model parameters.
• Refining the input-output relation of artificial muscles.
• Utilizing tracking performance as a metric for model parameter adjustment.

Main Results:

• The proposed controller achieves reference tracking with a root mean square error (RMSE) below 5%.
• Effective performance across varying stiffness levels and different load conditions.
• Demonstrated ability to capture soft muscle nonlinearity and adapt to changing environments.

Conclusions:

• The adaptive control method effectively models and controls soft artificial muscles.
• The approach enhances precision and adaptability in soft robotic applications.
• Validated effectiveness for wearable robots and operation in unstructured environments.