Denoising Diffusion Planner: Learning Complex Paths from Low-Quality Demonstrations

TL;DR

This paper introduces a diffusion-based path planning method that learns from low-quality demonstrations to generate complex, obstacle-avoiding paths for robotic manipulators, enhancing control with a diffusion model approach.

Contribution

It demonstrates that DDPMs trained on synthetic, low-quality data can generate complex robotic paths and integrates them into a receding-horizon control scheme.

Findings

Effective path generation from low-quality demonstrations

Successful obstacle avoidance in complex environments

Enhanced planning capabilities via diffusion models

Abstract

Denoising Diffusion Probabilistic Models (DDPMs) are powerful generative deep learning models that have been very successful at image generation, and, very recently, in path planning and control. In this paper, we investigate how to leverage the generalization and conditional sampling capabilities of DDPMs to generate complex paths for a robotic end effector. We show that training a DDPM with synthetic and low-quality demonstrations is sufficient for generating nontrivial paths reaching arbitrary targets and avoiding obstacles. Additionally, we investigate different strategies for conditional sampling combining classifier-free and classifier-guided approaches. Eventually, we deploy the DDPM in a receding-horizon control scheme to enhance its planning capabilities. The Denoising Diffusion Planner is experimentally validated through various experiments on a Franka Emika Panda robot.