GPU-oriented triangular grid collision detection method

Abstract

The invention discloses a GPU-oriented triangular grid collision detection method. The method includes a basic collision algorithm part and an algorithm optimization part. The basic collision algorithm part concretely includes the following steps: step 11, conducting a triangle intersection test; and step 12, conducting collision detection between two objects. The algorithm optimization part accelerates through a memory access reduction strategy and a memory sharing strategy, and concretely includes the following steps: step 21, optimizing a basic collision algorithm through a strategy of reducing the number of memory access; step 22, optimizing the basic collision algorithm through a strategy of memory sharing; and step 23, optimizing the basic collision algorithm through combination of reduction of the number of memory access and memory sharing. The GPU-oriented triangular grid collision detection method is suitable for models with medium complexity, reduces complexity of an algorithm, and improves the performance of a collision detection algorithm.

Description

technical field [0001] The invention relates to the field of computer graphics, in particular to a GPU-oriented triangular grid collision detection method. Background technique [0002] Collision Detection (CD for short) technology is used to detect whether there is overlap between one or more geometric objects in space, and if so, identify overlapping features. Collision detection algorithms are widely used in various applications in the field of computer graphics, including physics-based animation, geometric modeling, 6-DOF tactile rendering, robot path planning, medical imaging, interactive computer games, etc. [0003] With the rapid development of massively parallel processors such as multi-core graphics processing units (Graphics Processing Unit, referred to as GPU), and the general parallel computing architecture CUDA (Compute Unified Device Architecture, referred to as CUDA) launched by NVIDA. Many uniprocessor-based algorithms have been extended to the GPU using CU...

AI Technical Summary

Problems solved by technology

[0005] In view of the above-mentioned defects of the prior art, the technical problem to be solved by the present invention is to develop a GPU-oriented triangular mesh collision detection method, utilize the optimization of the basic collision detection algorithm to carry out large-scale parallel computing on the GPU to detect objects Perform fast and accurate collision detection. The algorithm is suitable for models of medium complexity. It involves methods such as triangle intersection testing based on separation axes, memory access optimization strategies, and shared memory optimization strategies, which reduce algorithm complexity and improve collision detection algorithms. performance

Images