Fast Online Optimization for Terrain-Blind Bipedal Robot Walking With a Decoupled Actuated SLIP Model

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• Bipedal robots require robust locomotion strategies for navigating complex, unpredictable environments.
• Existing methods often struggle with real-time adaptation to unobserved terrain variations and external disturbances.
• High-level motion planning and compliant foot design are crucial for enhanced stability and adaptability.

Purpose of the Study:

• To develop an online optimization algorithm for bipedal robot locomotion.
• To enable robots to blindly traverse diverse uneven terrains while maintaining stability.
• To enhance robot resilience against external pushes during locomotion.

Main Methods:

• Utilized a decoupled actuated spring-loaded inverted pendulum (aSLIP) model for motion planning.
• Formulated motion planning as a discrete-time model predictive control (MPC) problem solvable at 1 kHz.
• Integrated a whole-body controller with compliant feet for robust execution on the SLIDER robot.

Main Results:

• The robot successfully navigated various uneven terrains, including slopes, wave fields, and stairs, without prior knowledge.
• Demonstrated the ability to withstand pushes up to 40 N for 0.1 s while walking on uneven surfaces.
• Validated the effectiveness of the compliant feet in handling unobserved terrain variations.

Conclusions:

• The proposed online optimization algorithm significantly improves bipedal robot locomotion capabilities on challenging terrains.
• The decoupled aSLIP model and MPC approach provide an effective framework for real-time motion planning.
• Compliant foot design enhances robustness, paving the way for more versatile and reliable legged robots.