Autonomous Search & Navigation ROSBot
Autonomous exploration and hazard mapping system on ROS 2 Humble for the ROSBot platform. Maps unknown environments, localizes visual hazards in 3D space, and dynamically retraces its path home.
Completed RMIT Search & Navigation Challenge — autonomously exploring, mapping 5 hazards, and returning home within the 240s mission window.
Architected a custom 3-node ROS 2 stack (State Machine, Hazard Detection, Position Tracking) integrating Nav2, SLAM Toolbox, and explore_lite.
Engineered a robust FSM managing mission lifecycle with costmap-aware frontier selection and adaptive velocity.
Implemented continuous path-recording capturing waypoints during exploration for autonomous return to starting pose.
Built an RGB-D perception pipeline using OAK-D camera and find_object_2d to place hazard placards in the global map frame.
Designed a reactive spin-pause maneuver halting the robot on new hazard detection for accurate stationary depth samples.
3-node custom stack on standard AIIL navigation stack
RGB-D hazard localization with spin-pause depth sampling
Costmap-aware frontier exploration
Path-based autonomous return navigation
Adaptive velocity based on unexplored map area