Optimizing for what matters: The top grasp hypothesis

Abstract

In this paper, we consider the problem of robotic grasping of objects when only partial and noisy sensor data of the environment is available. We are specifically interested in the problem of reliably selecting the best hypothesis from a whole set. This is commonly the case when trying to grasp an object for which we can only observe a partial point cloud from one viewpoint through noisy sensors. There will be many possible ways to successfully grasp this object, and even more which will fail. We propose a supervised learning method that is trained with a ranking loss. This explicitly encourages that the top-ranked training grasp in a hypothesis set is also positively labeled. We show how we adapt the standard ranking loss to work with data that has binary labels and explain the benefits of this formulation. Additionally, we show how we can efficiently optimize this loss with stochastic gradient descent.In quantitative experiments, we show that we can outperform previous models by a large margin.

Robotics Dexterous Grasping: The Methods Based on Point Cloud and Deep Learning

“…Compared with traditional data-driven techniques, deep learning-based evaluation is more precise and robust. Kappler et al ( 2016 ) first indicates the feasibility of evaluating based on CNN. Inspired by this, Gualtieri et al ( 2016 ), Wang and Ling ( 2016 ), and Pas et al ( 2017 ) use LeNet (LeCun et al, 1998 ) to classify the grasp proposals and achieve an impressive performance.…”

High precision grasp pose detection in dense clutter

“…In Step 2 of Algorithm 1, we classify each candidate as a grasp or not using a four-layer convolutional neural network (CNN). CNNs have become the standard choice for GPD classification and ranking [8]. The main choice here is what CNN structure to use and how to encode to the CNN the geometry and appearance of the portion of the object to be grasped.…”