Robotic gas source localization and distribution mapping via deep reinforcement learning

Author

• Iliya KulbakaFollow

Year

2024

Season

Fall

Paper Type

Master's Thesis

College

College of Computing, Engineering & Construction

Degree Name

Master of Science in Computer and Information Sciences (MS)

Department

Computing

NACO controlled Corporate Body

University of North Florida. School of Computing

Committee Chairperson

Ayan Dutta, Ph.D.

Second Advisor

Swapnoneel Roy, Ph.D.

Rights Statement

http://rightsstatements.org/vocab/InC/1.0/

Third Advisor

O. Patrick Kreidl, Ph.D.

Department Chair

Zornitza Prodanoff, Ph.D.

College Dean

Dr. Klostermeyer

Abstract

This research aims to advance the fields of Gas Source Localization (GSL) and Gas Distribution Mapping (GDM) by developing deep reinforcement learning (DRL) methodologies suitable for complex, real-world environments. GSL and GDM are crucial for applications such as environmental monitoring, hazardous material detection, and search-and-rescue missions, where safe and efficient exploration is essential. Traditional methods often fall short in dynamic settings influenced by factors like wind and obstacles. To address these limitations, this study proposes novel neural network architectures and learning frameworks for adaptive exploration and mapping, integrating Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) layers, and Deep Q-Networks (DQN). The research explores the use of DRL in single robot scenarios to enhance real-time decision-making and operational efficiency. In GSL, the study focuses on accurately tracing gas sources by leveraging temporal and spatial data to guide the robot through complex environments. For GDM, it introduces a hybrid framework combining Gaussian Process Regression (GPR) and DRL to estimate gas concentrations from sparse samples, thereby reducing computational overhead during real-time deployment. Through extensive experimentation, the research demonstrates that the proposed methods can outperform traditional greedy and random walk-based strategies. Ultimately, this work seeks to contribute robust, adaptive exploration strategies for autonomous robotic systems in dynamic, hazardous, and computationally constrained environments.

Suggested Citation

Kulbaka, Iliya, "Robotic gas source localization and distribution mapping via deep reinforcement learning" (2024). UNF Graduate Theses and Dissertations. 1309.
https://digitalcommons.unf.edu/etd/1309