Multi-terrain traversability of mobile robots using deep learning

• Main Author: Lee, Jerome Zhi Hao
• Other Authors: Soong Boon Hee
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Electrical and electronic engineering::Computer hardware, software and systems
• Online Access: https://hdl.handle.net/10356/139599
• Institution: Nanyang Technological University

With robots being a growing topic in the world of technology, the need for more accuracy and precision is needed in its tasks. One such area is how these robots are becoming more able to navigate through various terrains to deliver and transport. In recent years, there has been a lot of projects and advancements explored to be used on such application. The possible functions that autonomous mobile robot navigation is limitless. With advancements in this research topic, we look to dig deeper at Deep Learning methods that can improve Traversability to manoeuvre any type of terrain. Various methods have been studied and put into effect as shown in recent advancements in delivery robots and the rise of autonomous self-driving cars. Even within deep learning methods, there are still much to discover as there are many factors to consider within the world of traversability due to various terrains and its unpredictability. In this project, the author aims to use a deep learning approach via convolutional neural networks to identify various terrains and use efficient algorithms for path planning. We will be using deep learning frameworks from TensorFlow and Keras, and input data from RGB images to help with terrain classification and path planning for multi-terrain traversability.