Back to search

West Virginia University

Imitation Learning in Robotic Manipulation Using Diffusion Models

Abstract

dc:description.abstract

This work deals with the problem of teaching robots by demonstration, also known by imitation learning. The main objective of this area of research is to enable the autonomous execution of complex robotic tasks using neural networks trained on expert-generated data, thus allowing the transfer of human knowledge to machines. To this end, this thesis describes an experimental setup especially designed for the study of imitation learning in robotic manipulation tasks and the adaptation and evaluation of a previously published technique to this setup. The experimental setup developed in this work is based on the Franka Emika Research 3 manipulator, which has seven degrees of freedom. This platform imposes strict communication requirements to ensure precision and safety, which demands a real-time Linux operating system. However, such systems often fail to recognize the Graphics Processing Unit (GPU) of a computer, which is an essential component for training and running neural networks. This work presents an integrated solution that ensures safe robot operation, including the sensing system (LiDAR and proprioception) and a teleoperation device based on a scaled replica of the manipulator, used to generate motion demonstrations. Given the recent and promising advances of Denoising Diffusion Probabilistic Models (DDPM) in robotic applications, this methodology was detailed and selected for practical implementation. The method leverages observations of both the environment and the robot itself to modulate a diffusion process that robustly generates actions, enabling the robot to perform complex tasks with consistent performance. Finally, the data collection procedure is described for a task designed to teach the robot to interact with an electrical cable, simulating a current measurement operation, an application with strong practical relevance for industry. The training procedures are also detailed, including data preprocessing and organization strategies, as well as the metrics used to evaluate the model’s performance. The results are discussed in light of the observed limitations, the robustness of the method, and the robot’s ability to generalize to previously unseen situations. The work concludes with a reflection on future perspectives, possible improvements to the experimental setup, and potential applications in other contexts.

Degree

thesis:*
Name thesis:degree_name
MS
Level thesis:degree_level
Thesis
Discipline thesis:degree_discipline
Mechanical and Aerospace Engineering

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Domingues de Oliveira, Marlon
Contributors dc:contributor
  • Guilherme Augusto Silva Pereira
  • Dimas Abreu Archanjo Dutra
  • Srinjoy Das

Subjects

dc:subject × 5

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
West Virginia University
Base URL
researchrepository.wvu.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Domingues de Oliveira, Marlon. Imitation Learning in Robotic Manipulation Using Diffusion Models. Thesis thesis, https://doi.org/10.33915/etd.13139