ALDM-Grasping: Diffusion-aided Zero-Shot Sim-to-Real Transfer for Robot Grasping

TL;DR

This paper introduces a diffusion-based framework that significantly improves zero-shot sim-to-real transfer for robotic grasping by enhancing simulation realism and reducing the reality gap, leading to higher success rates in diverse environments.

Contribution

The study presents ALDM, a novel diffusion model that enhances simulation fidelity for robotic grasping, enabling effective zero-shot transfer from simulation to real-world scenarios.

Findings

Achieves 75% success rate in simple backgrounds

Maintains 65% success rate in complex scenarios

Outperforms existing models in success and adaptability

Abstract

To tackle the "reality gap" encountered in Sim-to-Real transfer, this study proposes a diffusion-based framework that minimizes inconsistencies in grasping actions between the simulation settings and realistic environments. The process begins by training an adversarial supervision layout-to-image diffusion model(ALDM). Then, leverage the ALDM approach to enhance the simulation environment, rendering it with photorealistic fidelity, thereby optimizing robotic grasp task training. Experimental results indicate this framework outperforms existing models in both success rates and adaptability to new environments through improvements in the accuracy and reliability of visual grasping actions under a variety of conditions. Specifically, it achieves a 75\% success rate in grasping tasks under plain backgrounds and maintains a 65\% success rate in more complex scenarios. This performance demonstrates this framework excels at generating controlled image content based on text descriptions, identifying object grasp points, and demonstrating zero-shot learning in complex, unseen scenarios.