Linear dynamic separating axis algorithm for continuous collision detection

To address the issue of continuous collision detection's low detection efficiency, a fast continuous collision detection algorithm using linear dynamic separating axis is proposed to reduce the number of triangular elementary tests. Two separating axes were found before and after the triangle pair moved using an iterative algorithm, and they were then constructed as a linear dynamic separating axis. Finally, alternative culling algorithms or conventional methods were applied to the triangle pairs that could not be culled. The experimental results on several deformable models show that the elementary test culling rate of the proposed algorithm is over 89%, and it has higher detection efficiency in the narrow phase.