Files
Download Poster (498 KB)
Faculty Advisors
Satish Kumar, PhD
Description
Reinforcement Learning (RL), is a subfield of Machine Learning where a computational agent interacts with the environment, learning an optimal course of action by trial and error. Deep Reinforcement Learning (Deep RL) uses neural networks to learn to perform tasks directly from raw data, such as images or text, without hard-coding task-specific knowledge. In this context, datasets are collections of data used as a single unit for analytic and prediction purposes. Datasets are made for specific tasks with raw data specific to the task or machine being used. There is a need for increasingly robust datasets to increase the use and effectiveness of these tasks. The purpose of this work is to generate a dataset designed specifically for Opentrons Flex, a pipetting robot designed for high throughput and laboratory experiments. For this purpose, an attempt to generate a dataset using Opentrons API, and its protocols was done. Gazebo was utilized to simulate and acquire image data for Deep RL. However, this action was limited by the lack of documentation and files available to run gazebo simulations with. Opentrons API has no documentation that works to recreate its machines in gazebo’s virtual environment; without this, simulations for which data can be extracted cannot be done.
Publication Date
2024
Department
Computer Science
Recommended Citation
Davis, Lewis, "Generating a Dataset for Deep Reinforcement Learning" (2024). Undergraduate Research Posters 2024. 5.
https://engagedscholarship.csuohio.edu/u_poster_2024/5
Student Publication
This item is part of the McNair Scholars Program.
