Visually guided gait modifications for stepping over an obstacle: a bio-inspired approach

Area of Science:

• Robotics
• Artificial Intelligence
• Bio-inspired Engineering

Background:

• Robotic systems require autonomy and adaptability for unstructured environments.
• Animals provide bio-inspiration for robotic control structures, particularly learning for adaptability.

Purpose of the Study:

• To develop a system for quadruped robots to online learn obstacle detection and avoidance.
• To enable robots to adapt locomotion for safe navigation over obstacles.

Main Methods:

• Utilized a forward internal model to estimate perceptive information based on locomotion patterns.
• Employed a central pattern generator (CPG)-based locomotion controller for adaptive gait.
• Implemented online learning to adjust CPG parameters for stride length modification.

Main Results:

• The system successfully enabled a quadruped robot to learn to detect and avoid stumbling on obstacles.
• Online learning of locomotion adaptation and obstacle detection formed a sensorimotor map.
• Simulations demonstrated the feasibility of the proposed online learning approach.

Conclusions:

• The developed system allows quadruped robots to autonomously learn to navigate and overcome obstacles.
• Bio-inspired learning mechanisms enhance robotic adaptability in dynamic environments.
• This approach contributes to creating more autonomous and versatile robotic systems.