How to make a non lidar based map?

Hi Nvidia,

All of the isaac examples use lidar to create the map. However, we would like to create a map based on odometey measurements.

We want to have the user train the map boundary and exclusion areas.

What is the best way to achieve this within the isaac framework?

Isaac SDK 2020.2 wraps libraries GMapping and Cartographer for mapping. These use an occupancy grids from some depth sensor (lidar, stereo camera, etc.) and then register them together using odometry as a hint to create metrically accurate maps. Without the depth sensors, you would only have an estimate of a path with no map. Did you mean navigate on odometry only? (register initial position of robot in some pre-existing map and then localize on accumulate odometry only?)

Hi @hemals !! Happy Friday! So it sounds like I have to create my own mapping codelet and pass the grid points into Isaac as a lattice map? For our outdoor robots they use an RTK GPS solution which provides robot pose information in enough accuracy to be considered absolute position most of the time.