I am using ISAAC ROS Visual SLAM and Nvblox to produce a 3D map from a game field. All the measurement in game field is known and built. I can also successfully built the 3D map using nvblox by navigating around the game field with a mobile robot, but is there a way that I can evaluate the accuracy of the 3D map generated?
I know I can use Nav2 to evaluate, but I hope there is a way to compare the generated map with real world game field to know the accuracy of generated map. Since I am mapping a static game field, I have set the tsdf_decay_rate to 0, so the full map will be saved.
My environment and hardware:
Jetson Orin Nano 8GB with JetPack 6.1
Realsense D435i
ISAAC ROS packages Release 3.2
Hi @wongpeisen
We made different types of evaluations, speaking about the software: Performance Summary — isaac_ros_docs documentation
where in detail: isaac_ros_benchmark/benchmarks/isaac_ros_nvblox_benchmark/scripts/isaac_ros_nvblox_node.py at main · NVIDIA-ISAAC-ROS/isaac_ros_benchmark · GitHub
To run this benchmark: isaac_ros_benchmark — isaac_ros_docs documentation
If you are not interested in evaluating the sensor at the same time, I recommend using Isaac Sim to assess the simulation’s performance directly from your computer: Isaac Sim Examples — isaac_ros_docs documentation. This type of testing can help reduce noise and errors that may arise from different sensors and environments.
Best,
Raffaello