Autonomous Obstacle Crossing Strategies for the Hybrid Wheeled-Legged Robot Centauro

Area of Science:

• Robotics
• Artificial Intelligence
• Autonomous Systems

Background:

• Autonomous robots face challenges navigating unstructured environments.
• Hybrid legged/wheeled robots offer versatile mobility but require sophisticated control.
• Adapting locomotion to diverse terrain geometry is crucial for mission success.

Purpose of the Study:

• To introduce a methodology for autonomous terrain traversing in hybrid robots.
• To enable adaptable, extendable, and autonomously selected locomotion behaviors.
• To facilitate complex terrain negotiation through primitive behavior synthesis.

Main Methods:

• Utilizing a perception module with LiDAR point cloud data for terrain segmentation.
• Developing a set of adaptable terrain traversing primitive behaviors (driving, stepping).
• Implementing autonomous selection and regulation of behaviors based on terrain geometry.
• Serializing primitive behaviors to create complex terrain crossing plans.

Main Results:

• Successfully validated the methodology in Gazebo simulations.
• Demonstrated autonomous negotiation of diverse terrains on the CENTAURO robot.
• Showcased adaptable and extendable terrain traversing capabilities.

Conclusions:

• The proposed methodology enables autonomous, adaptable, and extendable terrain negotiation for hybrid robots.
• The system effectively utilizes environmental perception for behavior selection.
• The framework supports the integration of new behaviors for enhanced capabilities.