Metrics
42 File views/ downloads
189 Record Views
Thesis
Leader-Follower-Based Collaborative Navigation Of A UGV-UAV Formation
University of West Florida Libraries
Master of Science (MS), University of West Florida
2024
Abstract
This research study constructs and tests a leader-follower relationship between an unmanned aerial vehicle (UAV) and an unmanned ground vehicle (UGV) to determine its viability as a traversal method through complex environments. Given an environment with varying terrain, an aerial drone is a suitable candidate to perform coverage and run traversability analyses to build maps of navigable areas. This study seeks to take advantage of this ability, pairing the JetHexa (a hexapod UGV) with the CoDrone EDU (a quadrotor UAV) to create a dynamic in which the hexapod follows the leader UAV. In the course of this study, a convolutional neural network was used to create an object detection model via the You Only Look Once (YOLO) algorithm. This model was integrated into the UGV to identify and track the quadrotor. Custom, annotated images were combined with an online UAV data set to form the training set, and the training was performed using Google Colab. The model performs with confidence intervals ranging between 0.66 to 0.85, typically, but can reach values up to 0.95. To implement following behavior, two custom Robot Operating System (ROS) packages were developed. One, rosnp_msgs, is a message library that facilitates data transfer by converting Python Numpy arrays to transportable messages. The other, uav_follower, is a decoupled leader-follower system in which the UGV runs the developed object detection model in PyTorch to detect the UAV and subsequently uses its onboard depth camera to gather UAV distance data and, finally, calculate a target goal position. The resulting program is one in which the hexapod tracks a UAV with no inter-machine communication. Overall, in controlled environments, the hexapod UGV is able to follow the drone by navigating to derived target positions using its path planning system. The movement behavior resulting from said path planner, however, was deemed too advanced, leading the robot to traverse trajectories that were more complex than desired. It was determined that, given future improvements including a new global path planner and a more robust depth retrieval method, the leader-follower collaborative pair has potential in guiding a ground vehicle through rough terrain by outsourcing traversability analysis to the UAV.
Details
- Title
- Leader-Follower-Based Collaborative Navigation Of A UGV-UAV Formation
- Resource Type
- Thesis
- Contributors
- Hakki Erhan Sevil (Committee Chair)Yazan Alqudah (Committee Member)Tarek Youssef (Committee Member)
- Publisher
- University of West Florida Libraries
- Format
- pdf
- Number of pages
- 79
- Copyright
- © 2024 Terrance R. Williams, Jr.
- Identifiers
- 99380551696906600
- Academic Unit
- Dr. Muhammad Harunur Rashid Department of Electrical and Computer Engineering
- Language
- English
- Awarding Institution
- University of West Florida; Master of Science (MS)
- Theses and Dissertations
- Master of Science (MS), University of West Florida