I already have LunarSim running under ROS 2 Humble on Ubuntu 22.04.5 LTS and now need a plug-and-play execution layer that lets the simulated rover travel a user-defined distance in both X and Y while it autonomously skirts any rocks in its path. Key points of the job • Use the depth images published on /lunarsim/depth_image/depth as the sole sensing source. • Apply a classical computer-vision approach (OpenCV, PCL or similar) to detect rocks—no machine-learning training loops, just reliable depth-based segmentation. • When a rock is detected, perform a simple avoidance manoeuvre: veer around the obstacle, re-establish the original heading, and keep going. No global mapping or heavy path planning required. • Drive the rover exclusively through velocity commands so it behaves smoothly inside the simulator. Deliverables • A clean ROS 2 Humble package containing the navigation node(s), publishers/subscribers, and any helper libraries. • Launch file(s) that start the simulator interface, the obstacle-detection node, and the motion controller in one command. • Parameters for target X/Y distance, speed, safety margin, and timeout, all exposed through ROS 2 params. • README with build instructions (colcon), topic diagram, and a quick demo sequence I can reproduce. Acceptance test 1. Build and source the package in a fresh workspace. 2. Launch the system; specify e.g., X = 10 m, Y = 5 m. 3. Rover reaches the requested offset within a reasonable tolerance (<10 cm) without colliding with any rocks, issuing only velocity commands throughout the run. If something in my environment is unclear, let me know early—otherwise I look forward to integrating your package and watching the rover drive itself.