GPU-Accelerated Collision Analysis of Vehicles in a Point Cloud Environment.

Area of Science:

• Computer Science
• Robotics
• Computational Geometry

Background:

• Autonomous vehicle navigation in complex, enclosed environments presents significant challenges.
• Accurate and efficient collision detection is crucial for safe path planning.

Purpose of the Study:

• To present a GPU-accelerated collision detection method for vehicle navigation in enclosed spaces.
• To enable efficient path identification by analyzing vehicle-environment collisions using point cloud data.

Main Methods:

• A GPU-accelerated voxel Minkowski sum algorithm was developed for clearance analysis.
• Vehicle CAD models were converted to voxels for collision computation against environmental point clouds.
• Collision detection was performed adaptively without requiring mesh generation.

Main Results:

• The method efficiently computes collisions between vehicle voxels and environmental point clouds.
• Theoretical bounds for the accuracy of collision and clearance analysis were established.
• The GPU implementation was integrated with Unreal Engine for enhanced usability.

Conclusions:

• The proposed GPU-accelerated method provides an efficient solution for collision detection in vehicle navigation within enclosed spaces.
• The voxel-based approach offers an effective alternative to mesh-based methods for clearance analysis.
• Integration with Unreal Engine enhances the practical application and flexibility of the collision detection system.