Abstract: The existing collision detection algorithms for multi-robot are excessively time-consuming.To speed up the process, aparallel collision detection algorithm using dynamic bounding volume tree with a three-layer hierarchy was proposed.This new hierarchy consists of an upper dynamic removal layer, a middle link layer and a bottom layer.Firstly, the three-layer hierarchy was adopted to build a dynamic bounding volume tree between every two robots.Secondly, aparallel architecture was designed to compute each layer respectively and speed up the calculation in parallel.Lastly, parallel computing collision detection was implemented based on the three parallel structures using the OpenMP.Experiments performed on a 6/8 robots workstation and the analysis of time complexity show that under the same experimental conditions, when parallel computing the middle link layer and the bottom layer of the dynamic bounding volume tree, the efficiency of collision detection is about two times faster than that of the original data structure, and over four times faster than that of RAPID.The proposed parallel architecture fully takes advantage of the hierarchical structure of the dynamic bounding volume tree, and effectively improves the efficiency of collision detection among multi-robot.