A Novel Identification Methodology for the Coordinate Relationship between a 3D Vision System and a Legged Robot

Area of Science:

• Robotics
• Computer Vision
• Mechatronics

Background:

• Coordinate identification between robot vision systems and their environment is crucial for intelligent robotic applications.
• Current methods for this calibration suffer from limitations like complex feature matching, reliance on external equipment, and the need for multiple operators.

Purpose of the Study:

• To propose a novel, self-contained methodology for identifying the geometric parameters of 3D vision systems mounted on robots.
• To address the specific challenges associated with calibrating vision systems on legged robots.

Main Methods:

• The proposed method involves moving the robot on a relatively flat ground to estimate the ground plane and model the robot-ground relationship.
• The geometric parameters are identified by formulating the problem as an optimization task.
• The approach is designed to be performed without external tools or human intervention.

Main Results:

• The methodology was successfully integrated onto a legged robot named "Octopus".
• The calibrated "Octopus" robot demonstrated high stability in traversing rough terrains.
• Experimental results across diverse environments validated the accuracy and robustness of the proposed calibration method.

Conclusions:

• The novel methodology provides an effective and self-sufficient solution for calibrating robot vision systems.
• This advancement is particularly beneficial for legged robots, improving their environmental perception and navigation capabilities.
• The method offers a significant improvement over existing techniques, reducing complexity and resource requirements.