Robust Walking for Humanoid Robot Based on Divergent Component of Motion

Area of Science:

• Robotics
• Control Systems
• Biomechanics

Background:

• Humanoid robots require robust locomotion for complex tasks.
• Existing control methods face challenges with external disturbances and uneven terrain.

Purpose of the Study:

• To propose an improved model predictive control (MPC) method for robust humanoid robot walking.
• To enhance the robot's ability to handle disturbances and navigate uneven terrain.

Main Methods:

• Simplified humanoid robot model to a foot-pendulum model.
• Gait planning based on divergent components of motion (DCM) for single and double support phases.
• Model predictive control (MPC) with an extended Kalman filter (EKF) for DCM trajectory tracking.

Main Results:

• The proposed MPC-EKF controller effectively tracks the desired DCM trajectory.
• The controller compensates for center of mass (CoM) trajectory errors using step duration adjustment.
• Simulations demonstrate improved performance in disturbed and uneven terrain walking compared to traditional methods.

Conclusions:

• The improved MPC method based on DCM provides robust locomotion for humanoid robots.
• This approach offers enhanced stability and adaptability to environmental challenges.
• The method shows significant potential for real-world humanoid robot applications.