Optimizing actual PID control for walking quadruped soft robots using genetic algorithms

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• Developing effective control models and controllers for soft robots presents significant engineering challenges.
• Existing methods for soft robot locomotion often lack precision and efficiency.

Purpose of the Study:

• To propose and validate a new walking control method for quadruped soft robots using a genetic algorithm-optimized PID controller.
• To enhance the performance of soft robot locomotion through automated controller parameter tuning.

Main Methods:

• Constructed a control model correlating valve voltage with leg bending using geometrical analysis and novel sensor characteristics.
• Applied a genetic algorithm to automatically tune parameters and optimize Proportional-Integral-Derivative (PID) controllers.
• Demonstrated the method's application on a real 3D-printed quadruped soft robot for walking control.

Main Results:

• The genetic algorithm-optimized PID controller significantly improved trajectory tracking compared to the Ziegler-Nichols tuning method.
• Achieved an increase in walking speed from 5 mm/s to 8 mm/s, reduced error rate by 2.4064%, decreased overshoot by 12.55%, and shortened response time by 0.5 s.
• Outperformed particle swarm optimization by further reducing error rate by 0.4079%, overshoot by 8.4%, and response time by 1.0 s.

Conclusions:

• The proposed genetic algorithm-optimized PID method offers a superior approach for controlling quadruped soft robot locomotion.
• This optimization technique substantially enhances controller performance, leading to faster, more accurate, and stable robot movement.
• The study demonstrates a practical and effective solution for a key challenge in soft robotics.