【摘要】 针对航天员水下训练超大作业空间中的人机安全问题，提出了一种快速、精准的机械臂碰撞检测方法。首先，基于最小包络和二分法寻找模型最小包络球体及胶囊体，实现模型的最优球体及胶囊体包围盒的自动生成；然后，基于二叉树的思想，采用轴对齐包围盒(AABB)和球体及胶囊体包围盒2层结构，按照自顶向下的顺序对机械臂及舱体模型建立混合层次包围盒树，遵循同层优先的原则对层次包围盒树节点进行遍历，利用上层AABB层实现碰撞部位的快速定位，利用下层球体及胶囊体包围盒层保证碰撞检测的精度；最后，将该方法集成于自主研发的水下训练机械臂操控系统，进行了试验验证。结果表明：该方法减少了包围盒的多余体积，提高了碰撞检测的精度，计算效率提升了45%以上，满足超大作业空间碰撞检测准确性及实时性要求。更多还原

【Abstract】 Aiming at the man-machine safety problem of astronaut underwater training in the super large working space, a fast and accurate robot arm collision detection method was proposed. First, based on the minimum envelope and dichotomy, the sphere and capsule with the minimum envelope of the model were found, and the automatic generation of the optimal sphere and capsule bounding box of the model was realized. Then, based on the idea of binary tree, using the two-layer structure of axis aligned bounding box(AABB) and sphere and capsule bounding box, a hybrid hierarchical bounding box tree was established for the manipulator and cabin model in top-down order. The nodes of the hierarchical bounding box tree were traversed according to the principle of priority of the same layer. The collision location was realized by using the upper AABB layer. And the lower sphere and capsule bounding box layer were used to ensure the accuracy of collision detection. Finally, the method was integrated into the self-developed underwater training manipulator control system for verification. The experimental results showed that the redundant volume of the bounding box was reduced, the accuracy of collision detection was improved, and the calculation efficiency was improved by more than 45% which satisfied the requirements of accuracy and real-time of collision detection in large workspace.更多还原