We propose a novel approach to multifingered grasp planning that leverages learned deep neural network (DNN) models. We trained a voxel-based 3D convolutional neural network (CNN) to predict grasp-success probability as a function of both visual information of an object and grasp configuration. From this, we formulated grasp planning as inferring the grasp configuration that maximizes the probability of grasp success. In addition, we learned a prior over grasp configurations as a mixture-density network (MDN) conditioned on our voxel-based object representation. We show that this object-conditional prior improves grasp inference when used with the learned grasp success-prediction network compared to a learned, objectagnostic prior or an uninformed uniform prior. Our work is the first to directly plan high-quality multifingered grasps in configuration space using a DNN without the need of an external planner. We validated our inference method by performing multifinger grasping on a physical robot. Our experimental results show that our planning method outperforms existing grasp-planning methods for neural networks (NNs).

中文翻译：

---

在深层神经网络中通过推理进行多指抓取计划：通过学习可区分的模型胜过抽样

我们提出了一种新颖的方法来进行多指抓取计划，该方法利用了学习的深度神经网络（DNN）模型。我们训练了基于体素的3D卷积神经网络（CNN），以根据物体的视觉信息和抓握配置预测抓握成功概率。据此，我们制定了抓地计划，以推断可最大程度提高抓地成功概率的抓地配置。此外，我们还学习了基于混合体密度网络（MDN）的先入为主的配置，该网络以基于体素的对象表示为条件。我们显示，与学习的，对象不可知的先验或不知情的先验先验相比，该对象条件先验与学习的学习成功预测网络一起使用时，可改善抓取推理。我们的工作是第一个使用DNN直接在配置空间中直接计划高质量多指握把而无需外部计划者的工作。我们通过在物理机器人上执行多指抓握来验证我们的推理方法。我们的实验结果表明，我们的计划方法优于现有的神经网络（NN）把握计划方法。