Core Concepts
An autonomous robotic system using the Turtlebot3 Burger and ROS Noetic that can generate a comprehensive map of unknown environments and accurately locate and estimate the poses of "victims" represented by AprilTags.
Abstract
The research article presents the design and implementation of an autonomous robot system using the Turtlebot3 Burger (TB3) and Robot Operating System (ROS) Noetic. The system aims to address the critical need for effective reconnaissance in disaster scenarios by generating a comprehensive map of unknown environments and identifying any present "victims" using AprilTags as stand-ins.
The key components of the system include:
Hardware Setup: The TB3 platform was selected due to its modularity, ease of use, and compatibility with ROS. The system was configured with a 360-degree LiDAR sensor, a Raspberry Pi camera, and additional hardware like an external battery pack and Wi-Fi adapter to enhance its capabilities.
Software Setup: The system utilizes existing ROS packages like explore_lite for frontier-based exploration, move_base for navigation, and apriltag_ros for AprilTag detection, as well as custom nodes like ckf3D for recursive Bayesian estimation of AprilTag positions and search_and_rescue for comprehensive search of the mapped environment.
Exploration Algorithm: The authors implemented a more efficient and effective exploration algorithm that combines frontier-based and next-best-view approaches. It uses an expanding wavefront frontier detection algorithm and computes exploration goals by sampling around free space to maximize information gain.
AprilTag Pose Estimation: To address the bias in the apriltag_ros package's position estimates, the authors implemented a Cubature Kalman Filter (CKF) to improve the accuracy of AprilTag localization.
Search and Rescue: After the exploration phase, the system employs a grid-based decomposition of the environment to plan a zig-zag search pattern, ensuring comprehensive coverage and detection of all AprilTags.
The authors evaluated the system's performance in both simulated Gazebo environments and a real-world arena, demonstrating its ability to accurately map the environment and locate the majority of the AprilTags. The CKF-based pose estimation was shown to significantly improve the accuracy of AprilTag localization compared to the apriltag_ros package.
The article also discusses the lessons learned from various challenges encountered, such as drift issues, carpet-related problems, and hardware failures, providing valuable insights for future development and deployment of such autonomous robotic systems.
Stats
The average Mean Squared Error (MSE) for the AprilTag position estimates using the CKF was 0.15 meters in the TB3 World arena and 0.30 meters in the House arena, compared to 0.27 meters and 0.36 meters, respectively, using the apriltag_ros package.
Quotes
"Just like turtles, our system takes it slow and steady, but when it's time to save the day, it moves at ninja-like speed!"
"Despite Donatello's shell, he's no slowpoke - he zips through obstacles with the agility of a teenage mutant ninja turtle."