TY - UNPB
T1 - Denoising Diffusion Planner
T2 - Learning Complex Paths from Low-Quality Demonstrations
AU - Nikken, Michiel
AU - Botteghi, Nicolò
AU - Roozing, Wesley
AU - Califano, Federico
PY - 2024/10/28
Y1 - 2024/10/28
N2 - 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.
AB - 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.
KW - cs.RO
U2 - 10.48550/arXiv.2410.21497
DO - 10.48550/arXiv.2410.21497
M3 - Working paper
BT - Denoising Diffusion Planner
PB - ArXiv.org
ER -