Hierarchical Policies for Cluttered-Scene Grasping with Latent Plans

TL;DR

This paper introduces a hierarchical, end-to-end learning framework for 6D robotic grasping in cluttered scenes, leveraging latent plan representations and reinforcement learning to improve collision avoidance and generalization.

Contribution

It presents a novel hierarchical approach combining latent plan encoding, variational autoencoders, and reinforcement learning for robust cluttered-scene grasping.

Findings

Outperforms baselines in simulation

Generalizes to real-world cluttered scenes

Effective in complex, obstacle-rich environments

Abstract

6D grasping in cluttered scenes is a longstanding problem in robotic manipulation. Open-loop manipulation pipelines may fail due to inaccurate state estimation, while most end-to-end grasping methods have not yet scaled to complex scenes with obstacles. In this work, we propose a new method for end-to-end learning of 6D grasping in cluttered scenes. Our hierarchical framework learns collision-free target-driven grasping based on partial point cloud observations. We learn an embedding space to encode expert grasping plans during training and a variational autoencoder to sample diverse grasping trajectories at test time. Furthermore, we train a critic network for plan selection and an option classifier for switching to an instance grasping policy through hierarchical reinforcement learning. We evaluate our method and compare against several baselines in simulation, as well as demonstrate that our latent planning can generalize to real-world cluttered-scene grasping tasks. Our videos and code can be found at https://sites.google.com/view/latent-grasping .