Objective
Collaborated in a team to develop an Autonomous Reconnaissance Robot on the TurtleBot3 platform, designed to map and search for victims in disaster-stricken areas.
Approach
Victim Detection -- Simulated victims using AprilTags, detected with a monocular camera and OpenCV. Subsequently enhanced the system by integrating object detection functionality to identify real people.
​
Navigation -- Implemented SLAM using Gmapping, and autonomous navigation using move_base ROS package.
​
Exploration -- Integrated explore_lite ROS package with the navigation stack to autonomously explore the environment.
The baseline approach initially struggled to detect all victims. To address this, we implemented three goal-focused exploration strategies, one of which achieved a significant 90% improvement in victim detection.
Personal Contribution
-
Team lead
-
Integrated the hardware components on the robot
-
​Implemented the entire navigation and exploration stack