Stereo mapping with odometry from only IMU, how to

[alexr]

Hi Mathieu,

I need to know whether I can fuse IMU data to generate odometry for the mobile robot. Our mobile robot does not have a built in encoder and we typically use IMU to compensate for it.

The idea is to use the robot outside for agricultural applications using stereo camera. The robot is relatively bulky (small tractor size).

1. Can you guide me as to how to do that using RTABMAP (I plan to use ROS for IMU and stereo data without robot odometry. The IMU is relatively expensive and accurate enough)
2. Considering the robot will be traversing on uneven surface, will RTABMAP be able to map the terrain properly.
3. As for the stereo, does Zed stereo camera provide decent map/ point cloud to utilize for later data processing (can you suggest some other stereo camera that will be suitable for outside use).

Thanks

Alex

[matlabbe]

Hi Alex,

1. I recently made an example launch file (sensor_fusion.launch) using robot_localization package to do fusion of IMU + Visual odometry with an extended Kalman filter (EKF). You can use this example to get you started.
2. RTAB-Map can map in 6DoF, so there is no problem to map a 3D terrain.
3. Yes, the ZED (and other stereo cameras) can be a good sensor for your outdoor application. Depending on the range precision you need, having two synchronized cameras with custom baseline can also be useful. Depending on your budget, you can also think about a Velodyne sensor to create even more precise point clouds.

cheers,
Mathieu