Machine learning for control of robotic arms

Robots are seen as a promising tool to help humans improve their productivity and living standard, with the advancement of technology, the introduction of machine learning has led to new developments in robotic arm control solutions. There are several mainstream machine learning control robotic arm...

Full description

Saved in:
Bibliographic Details
Main Author: Wu, Xiaoyang
Other Authors: Wen Bihan
Format: Thesis-Master by Coursework
Language:English
Published: Nanyang Technological University 2023
Subjects:
Online Access:https://hdl.handle.net/10356/165263
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Robots are seen as a promising tool to help humans improve their productivity and living standard, with the advancement of technology, the introduction of machine learning has led to new developments in robotic arm control solutions. There are several mainstream machine learning control robotic arm solutions such as learning from demonstration and reinforcement learning. This dissertation compares their respective advantages and disadvantages. Reinforcement learning-based algorithms can better face complex unknown scenarios and can be combined with migration learning to achieve a smooth transition of the model from the simulated environment to the real world. In this dissertation, based on a model-free deep reinforcement learning algorithm, we illustrate the related concepts involved and then verify its performance under light noise or irregular objects through experiments.