Evaluating Gaussian Grasp Maps for Generative Grasping Models

TL;DR

This paper introduces a continuous Gaussian grasp map representation for training generative grasping models, leading to higher success rates and better real-world transferability compared to traditional binary maps.

Contribution

It proposes a Gaussian representation for grasp annotations, improving training data quality and success rates in simulated and real robotic grasping tasks.

Findings

Gaussian maps outperform binary maps in success rates

Models trained with Gaussian maps transfer effectively to real robots

Achieved 87.94% accuracy in simulated grasping benchmark

Abstract

Generalising robotic grasping to previously unseen objects is a key task in general robotic manipulation. The current method for training many antipodal generative grasping models rely on a binary ground truth grasp map generated from the centre thirds of correctly labelled grasp rectangles. However, these binary maps do not accurately reflect the positions in which a robotic arm can correctly grasp a given object. We propose a continuous Gaussian representation of annotated grasps to generate ground truth training data which achieves a higher success rate on a simulated robotic grasping benchmark. Three modern generative grasping networks are trained with either binary or Gaussian grasp maps, along with recent advancements from the robotic grasping literature, such as discretisation of grasp angles into bins and an attentional loss function. Despite negligible difference according to the standard rectangle metric, Gaussian maps better reproduce the training data and therefore improve success rates when tested on the same simulated robot arm by avoiding collisions with the object: achieving 87.94\% accuracy. Furthermore, the best performing model is shown to operate with a high success rate when transferred to a real robotic arm, at high inference speeds, without the need for transfer learning. The system is then shown to be capable of performing grasps on an antagonistic physical object dataset benchmark.