ROS SLAM: 2D & 3D Mapping

Job ID: 39799156

Budget: ₹1,500 – ₹12,500 INR

A ROS-based SLAM pipeline must be assembled that can generate both 2D occupancy grids and 3D point-cloud maps from a stereo-camera feed. The system will run on ROS (Noetic or Humble—whichever you recommend for best library support) and should operate in real time on typical mobile-robot hardware.

I will supply the raw stereo images and calibration files; you turn them into an integrated SLAM stack that launches from a single ROS launch file, subscribes to the stereo topics, performs feature extraction, state estimation, loop closure, and then publishes:

• /map → nav_msgs/OccupancyGrid (2D)
• /map_3d → sensor_msgs/PointCloud2

Accuracy within ±2 cm over a 10 m loop, and a drift rate below 1 % of distance travelled, will be used as acceptance criteria. The final package must include:

• Source code with clear comments
• ROS launch and parameter files
• Short README describing setup, dependencies, and how to record/visualise results in RViz
• Brief performance report validating the metrics above on my log bag

Open-source libraries such as RTAB-Map, ORB-SLAM3, GTSAM, OpenCV, or PCL are welcome as long as the final solution remains ROS-compatible and buildable with catkin/colcon.

Timeline is flexible, but an initial demo video showing live mapping in RViz within two weeks would be ideal.