Some obstacles/boundaries are not detectable via LIDAR or camera PointCloud but are known through the CV team's object detection (eg. lane boundaries). In order for the Navigation Stack to utilize them, we need to artificially create them so they can be avoided. We most likely need to modify either the costmap(s) or occupancy grid but we don't know which...
Update: @gmin7 and the UTRA ART CV team have a lane detection package. We need to aid in integrating it into our code base, with the Navigation Stack and our SLAM package. A video feed is needed for the Lane Detection package
Goals:
[ ] Determine how the Costmap/Occupancy Grid messages are structured
[ ] Determine how to edit Costmap/Occupancy Grid as to introduce artificial obstacles
[x] Create nodes (Lane Detection package) that enable to programmatically edit the Costmap/Occupancy Grid
[x] Feed in edited Costmap/Occupancy Grid to the Navigation Stack
[ ] Provide video feed to Lane Detection package
Acceptance Criteria:
[ ] Wiki explaining Costmap and Occupancy Grid structure and how to edit them
[x] Edited Costmap/Occupancy Grid can be visualized in rviz
[ ] Navigating robot will navigate around artificial obstacles
Notes:
Wiki should explain which of Costmap or Occupancy Grid should be modified to introduce artificial obstacles and why.
How to edit the "maps" should also include how to convert a meter value to an index value in however the map is stored. Eg. which index does 3 meters in front of the robot represent?
Learning how to remap a ROS topic will be of use when re-routing output topics for the "map" processing nodes.
Description:
Goals:
Acceptance Criteria:
Notes:
remap
a ROS topic will be of use when re-routing output topics for the "map" processing nodes.