A Framework for Multiple Ground Target Finding and Inspection Using a Multirotor UAS.

Area of Science:

• Robotics
• Computer Vision
• Aerospace Engineering

Background:

• Small unmanned aerial systems (UASs) possess advanced navigation, enabling new data collection methods.
• Applications like inspection, search and rescue, and precision agriculture require target identification and interaction capabilities.
• Low-cost UAS sensors and external disturbances present challenges for accurate target localization and action.

Purpose of the Study:

• To propose a robust framework for vision-based ground target finding and action using the Observe, Orient, Decide and Act (OODA) decision-making approach.
• To implement this framework as a modular software system within the Robotic Operating System (ROS).
• To demonstrate the framework's effectiveness in applications requiring single or multiple target detection and action.

Main Methods:

• Development of a vision-based framework utilizing the OODA loop for decision-making.
• Implementation of the framework as a modular software system in ROS.
• Testing the framework's robustness against localization and target detection errors through two distinct test cases.

Main Results:

• The implemented framework demonstrated robustness against localization and target detection errors.
• The system successfully performed tasks involving ground target finding and subsequent actions.
• The framework proved effective in handling challenges posed by low-cost sensors and external disturbances like wind.

Conclusions:

• The proposed OODA-based framework provides a robust solution for vision-based target finding and action in UAS applications.
• The modular ROS implementation allows for flexible deployment across various scenarios.
• The system's ability to overcome sensor inaccuracies and environmental disturbances highlights its practical utility.