To perform simultaneous mapping and localization (SLAM), we use the following packages:
- OctoMap - 3D occupancy grid mapping approach;
- RTAB-Map - Visual odometry (F2M, using GFTT and ORB features) and loop closure, integrating OctoMap mapping;
- robot_localization - State estimation node using Unscented Kalman Filter (UFK) to fuse Visual Odometry with IMU data;
This package contains the launch files to run the packages automatically. To start the SLAM process, use the following slam-navigation.launch. You can use the parameters with_camera to enable the camera automatically, and with_path_finding to start the path finding node, like this:
roslaunch mapping slam-navigation.launch with_camera:=true with_path_finding:=true
The tests were measured with the pre-recorded Test4_*.bag files at 30 FPS. The table shows the average Rate of odometry and map update to each testbench.
| Testbench | Odometry Rate (Hz) | Map Update Rate (Hz) |
|---|---|---|
| CPU: Ryzen 3700x, RAM:16GB, GPU: RTX3070 | 30.0 | 14.3 |
| Jetson Xavier Nx (8GB) | 8.77 | 7.65 |