Robotic Online Path Planning on Point Cloud.

Area of Science:

• Robotics
• Computer Vision
• Computational Geometry

Background:

• Path-planning for mobile robots in 2.5-D environments presents challenges due to complex terrain.
• Traditional methods often rely on computationally expensive surface reconstruction or meshing.

Purpose of the Study:

• To develop an efficient and robust path-planning method for mobile robots in 2.5-D environments using raw point cloud data.
• To avoid complex surface parametrization and reconstruction techniques.

Main Methods:

• Utilizes the 3-D tensor voting framework on raw point clouds for surface analysis.
• Employs a novel local Riemannian metric based on saliency components derived from tensor voting.
• Leverages graphics computation units for accelerated tensor voting computations.

Main Results:

• Demonstrates that geodesic paths in the 3-D tensor space yield effective path-planning results.
• The proposed method achieves smoother robot maneuvers compared to traditional approaches.
• Maintains minimal travel distance while improving maneuverability.

Conclusions:

• The novel Riemannian metric and tensor voting approach offer an efficient solution for 2.5-D robot navigation.
• This method provides a computationally advantageous alternative to traditional surface reconstruction techniques.
• The findings contribute to advancing autonomous robot navigation in complex, real-world environments.