Controlling of Pneumatic Muscle Actuator Systems by Parallel Structure of Neural Network and Proportional Controllers...

Area of Science:

• Robotics and Control Systems
• Artificial Intelligence in Engineering

Background:

• Pneumatic Muscle Actuators (PMAs) exhibit complex non-linear dynamics, posing challenges for precise control.
• Existing control methods often struggle to accurately capture and compensate for these non-linearities, limiting performance in applications.

Purpose of the Study:

• To introduce a novel Parallel Neural Network Proportional (PNNP) controller designed to accurately track the non-linear behavior of PMAs.
• To evaluate the PNNP controller's effectiveness in controlling both elongation and bending motions of different PMA types under varying load conditions.

Main Methods:

• A hybrid controller architecture combining a neural network (NN) with a proportional (P) controller was developed.
• The PNNP controller was implemented and tested on a single extensor PMA and a single self-bending contraction actuator (SBCA).
• Performance was assessed across various load conditions, and the controller was further validated in a human-robot shared control system.

Main Results:

• The PNNP controller demonstrated a high level of precision and fast-tracking capabilities for PMA control.
• Effective control of both elongation (extensor PMA) and bending angle (SBCA) was achieved under different load values.
• Successful application in a human-robot shared control system highlighted the controller's practical utility and efficiency.

Conclusions:

• The proposed PNNP controller structure is highly effective for precisely controlling the non-linear dynamics of pneumatic muscle actuators.
• The PNNP controller offers a robust and efficient solution for advanced robotic applications requiring accurate and responsive actuation.
• This novel approach advances the state-of-the-art in PMA control, enabling more sophisticated human-robot interaction and automation.