An effective method for grasp planning on objects with complex geometry combining human experience and analytical approach

Abstract

In this paper, an effective method for identifying the graspable components of objects with complex geometry is proposed for grasp planning based on human experience. Instead of focusing on individual objects, our method identifies graspable components on the category level under the assumption that geometrically alike objects share similar graspable components. Firstly, employing a modified SHOT descriptor, a fast KNN-based method is developed for object categorization. Then, the graspable components are identified by adopting a learning framework based on human experience. Afterwards, a fast analytical grasp planning method is proposed which comprises of contact points exaction and hand kinematics calculation. Finally, a regression model based on the extreme learning method (ELM) is built which inputs the desired contact points and the wrist orientation and outputs the wrist position. This approach is time-saving comparing with the optimization method. The simulations and experiments demonstrate the effectiveness of the proposed approach by realizing grasps on the graspable components of human choice for objects with complex geometry.

创新点

本文提出一种有效的基于人类经验识别复杂形状物体可抓取部位的方法。假设形状相似物体(功能也相似)的可抓取部位也相似, 所提方法能够在物体类别层面上识别可抓取部位。首先, 算法对SHOT三维形状描述子进行改进, 提出一种基于KNN的物体分类方法, 并实现可抓取部位的识别; 在此基础上, 提出一种快速的抓取规划方法, 包括力封闭点的提取与手运动学求解; 最后, 利用极限学习机(ELM), 建立抓取点、手腕方向与位置之间的回归模型。这种方法比运动学求解的方法更节省时间。所提算法的有效性在仿真与机器人平台上得到验证。