ethzasl elevation_mapping behaviour

asked 2018-07-10 10:35:42 -0500

teonik gravatar image

I am using ethzasl elevation_mapping to build a height map from lidar pointclouds. We're running in some problems, and I would appreciate if someone could help us out:

  1. I can't find any documentation on the package, save for the brief Github description. For such a complex piece of mathematical machinery, trial and error doesn't work good. For instance, I couldn't find any docs for the save_map service and had to guess its existence. Gentlemen, where are the docs? Are there any?
  2. When we run the package and feed our data in it, it spams warnings about being unable to find "/sensor" frame in tf. Our pointclouds are in the "/Sensor" frame, and it is specified in the configs. Adding a static all-zeros publisher from /Sensor to /sensor seems to solve the problem, but it is weird.
  3. On our robot we have a pretty accurate XY-yaw positioning, and an IMU for roll/pitch. The IMU seems to be slightly out of synch at some times, so we get phantom waves on the ground. Does the package provide some instance of pointcloud matching? If yes, how can we control its work?
  4. What is the importance of the /pose topic input being published separately from the tf tree? In our tf transform to sensor frame the XY-yaw is updated roughly at 100 Hz, and IMU roll-pitch transform is updated at about 400 Hz. We synthesize the necessary /pose at 100 Hz, may that cause problems?
  5. The position_x and position_y parameters seem to change nothing: the map looks centered at (0,0). E.g. when we specify length_in_x: 500.0, the map cuts off at x=-250, same way with position_x: -250.0 or position_x: 0.0. Both parameters are loaded from a YAML file which is modified in numbers from long_range.yaml present in the package.

Sincerely grateful for any answers.

edit retag flag offensive close merge delete