Experimental Investigations into Using Motion Capture State Feedback for Real-Time Control of a Humanoid Robot

Area of Science:

• Robotics
• Control Systems
• Humanoid Locomotion

Background:

• Humanoid robots struggle with dynamic locomotion tasks.
• Key challenges include motion planning, feedback control, and state estimation.
• Internal state estimation can be unreliable for complex systems.

Purpose of the Study:

• To investigate using external motion capture for state feedback in humanoid robot control.
• To compare motion capture feedback with internal state estimation.
• To assess the feasibility of motion capture in high-frequency feedback loops.

Main Methods:

• Utilized an external motion capture system for state feedback.
• Implemented an online whole-body controller for the RH5 humanoid robot.
• Compared motion capture feedback against an internal state estimator (IMU, kinematics, contact sensing).
• Tested motions included squatting and single-leg balancing.

Main Results:

• External motion capture systems can be integrated into high-frequency feedback control loops.
• Motion capture provided reliable state feedback for complex locomotion tasks.
• Demonstrated successful execution of squatting and single-leg balancing using motion capture feedback.

Conclusions:

• State-of-the-art motion capture systems offer a viable alternative for humanoid robot state feedback.
• Motion capture can improve the reliability of control in challenging locomotion scenarios.
• This approach addresses limitations of internal state estimation in humanoid robotics.