Grasp detection for assistive robotic manipulation

Abstract

In this paper, we present a novel grasp detection algorithm targeted towards assistive robotic manipulation systems. We consider the problem of detecting robotic grasps using only the raw point cloud depth data of a scene containing unknown objects, and apply a geometric approach that categorizes objects into geometric shape primitives based on an analysis of local surface properties. Grasps are detected without a priori models, and the approach can generalize to any number of novel objects that fall within the shape primitive categories. Our approach generates multiple candidate object grasps, which moreover are semantically meaningful and similar to what a human would generate when teleoperating the robot—and thus should be suitable manipulation goals for assistive robotic systems. An evaluation of our algorithm on 30 household objects includes a pilot user study, confirms the robustness of the detected grasps and was conducted in real-world experiments using an assistive robotic arm.

Fast geometry-based computation of grasping points on three-dimensional point clouds

“…The authors then matched these parts against template grasps. Similarly, Jain and Argall 20 described an algorithm to match whole real objects against geometric shapes (i.e. spheres, cylinders, boxes) rather than parts of them.…”

From Single 2D Depth Image to Gripper 6D Pose Estimation: A Fast and Robust Algorithm for Grabbing Objects in Cluttered Scenes

“…The authors in [18] propose an approach that takes 3D point-cloud and hand geometric parameters as the input, then search for grasp configurations within a lower dimensional space satisfying defined geometric necessary conditions. Jain et al [19] analyze the surface of every observed point cloud cluster and automatically fit spherical, cylindrical, or box-like shape primitives to them. The method uses a predefined strategy to grasp each shape primitive.…”

Grasping Unknown Objects Based on Gripper Workspace Spheres

“…Object grasping problem is usually approached based on physical analysis (classical approach) [1]- [4], geometry [5]- [8], or machine learning (ML) [9]- [11]. The first requires sufficient knowledge about the object (shape, mass, material, ...…”

“…Their algorithm computes grasping points in lower than 1 second duration, however is limited to 2 fingered grippers, and no data regarding grasping success rate is presented. Grasp planning of unknown objects from point cloud data is presented in [5], using geometric information to categorize objects into shape primitives, with predefined strategies for each. Success rate of 82% is achieved, however no computation time data was reported.…”