English

Learning Probabilistic Multi-Modal Actor Models for Vision-Based Robotic Grasping

Robotics 2019-04-17 v1 Machine Learning

Abstract

Many previous works approach vision-based robotic grasping by training a value network that evaluates grasp proposals. These approaches require an optimization process at run-time to infer the best action from the value network. As a result, the inference time grows exponentially as the dimension of action space increases. We propose an alternative method, by directly training a neural density model to approximate the conditional distribution of successful grasp poses from the input images. We construct a neural network that combines Gaussian mixture and normalizing flows, which is able to represent multi-modal, complex probability distributions. We demonstrate on both simulation and real robot that the proposed actor model achieves similar performance compared to the value network using the Cross-Entropy Method (CEM) for inference, on top-down grasping with a 4 dimensional action space. Our actor model reduces the inference time by 3 times compared to the state-of-the-art CEM method. We believe that actor models will play an important role when scaling up these approaches to higher dimensional action spaces.

Keywords

Cite

@article{arxiv.1904.07319,
  title  = {Learning Probabilistic Multi-Modal Actor Models for Vision-Based Robotic Grasping},
  author = {Mengyuan Yan and Adrian Li and Mrinal Kalakrishnan and Peter Pastor},
  journal= {arXiv preprint arXiv:1904.07319},
  year   = {2019}
}
R2 v1 2026-06-23T08:40:26.901Z