Tactile identification of textures using machine learning/ deep learning

Abstract

In the recent years, Artificial Intelligence technology has grown exponentially, especially in sub-areas such as Machine Learning and Deep Learning. However, many of its applications rely on image acquisition techniques to obtain datasets. This project seeks to establish a method to identify an object, based on its haptic properties, and using Machine Learning and/or Deep Learning techniques.

In this project, different materials are tested across a consistent laboratory setup by applying the same amount of force over time through the use of a robotic arm in a laboratory setup. The pressure received on each material is recorded as a dataset. The collected datasets are then fed into a coded program containing a Machine Learning/ Deep Learning algorithm, where the algorithm learns the characteristics of each labelled dataset and establishes a trained model based on the set of material and algorithm. Thereafter, the unseen dataset of a material can be tested on the newly trained Machine Learning/ Deep Learning model, to predict and identify the material as a result.

This report will cover on the considerations behind the project, the documentation of experimental processes and parameters, and provide an analysis and conclusion for the above.