GOFIRST-Robotics / NASA-RMC-2020-NorthstarRobotics

The NASA RMC 2020 repository for Northstar Robotics, a segment of University of Minnesota Robotics
https://docs.google.com/presentation/d/1eXhP5dbxtgWJBE492b78TFrnQFL9yoPsJFO4HszJn60/edit?usp=sharing
Other
3 stars 0 forks source link

Mapping #32

Open libBAS opened 4 years ago

libBAS commented 4 years ago

Get mapping working (see Matt)

juliaschatz commented 4 years ago

Interesting things: http://wiki.ros.org/rtabmap_ros

juliaschatz commented 4 years ago

How do we go from identified map -> costmap?

libBAS commented 4 years ago

What do you mean by identified map? Like what is it/what format?

JWCS commented 4 years ago

(For RMC, rtabmap isn't useful, b/c the map is already fully defined. Obstacles don't count, b/c that's the purpose of the occupancy grid layer of the costmap) Rtabmap can publish the tf transform from map->odom (localization), and an OccupancyGrid (a layer to) the costmap (as the 'map' topic). See also this example launch file: https://github.com/introlab/rtabmap_ros/blob/master/launch/demo/demo_turtlebot3_navigation.launch

juliaschatz commented 4 years ago

Can you suggest an alternate workflow to establish the occupancy grid?

JWCS commented 4 years ago

https://wiki.ros.org/depthimage_to_laserscan https://wiki.ros.org/pointcloud_to_laserscan image In this image, from move_base, the odom (plus frame tf's) from the ekf (possibly seeded with an absolute position estimate such as gps, decawaves, or amcl), plus a map (from a static map def, since we know the map), and a laser_scan are enough (plus goal).

JWCS commented 4 years ago

Even a pointcloud works, I guess, so immediate conversion needed when the realsense publishes a pointcloud. But if you're using amcl to test on the conference room for localization, then you'll need a laser_scan