Thursdays 3:00 pm Pacific Time - No meeting Nov 28th (US Holiday)
Link to meeting on ROS2 Calendar: Navigation2 WG
To receive invites, add yourself to the google group: firstname.lastname@example.org
Nov 14 2019 Notes
Attendees: Geoff Biggs, Brian Wilcox , Carl Delsey, Carlos Orduno, Steve Macenski, Shivang Patel, Matt Hansen, Yathartha Tuladhar, Mohammad Haghighipanah, Melih Erdogan, Francisco Rico, John Connolly
#1292 - Controller PR from Steve Macenski needs new reviewer - Carlos will look into it.
#1343 Travis build is failing
- Carl is talking a look at it. It is timing out (even without the tests).
They had same issues in Autoware (Geoff) - could ask for longer timeout, Autoware was increased from 50 minutes to 100.
Test failing even when just building (Carl). It could be Gazebo ros plugins.
(Carlos) Multi behavior nodes add build time.
(Carl) Travis is probably a second/third opinion of whether things are working.
(Carl) It will take effort to make it passing. Ruffin maintains Circle, so we can just use Circle.
- Option 1: try to get more time.
- Option 2: try to get the time down. reduce the number of tests. Multi-robot test takes a lot of time.
- Option 3: remove Travis.
VOTE: Disable Travis. Put effort into Circle CI.
#808 Waypoint Following:
- (Matt) First come up with use cases. E.g., agricultural use cases where it has to go through different points to complete a task.Another example is a patrol robot with geofenced area.
(Mohammad) Waypoint following should be in different repo.
(Steve) neutral; Most people use it in warehouse; Lets not put complexity in deeper levels; Option 1 favorable.
(Matt) Could this be a performance bottleneck if we keep everything very separate.
(Steve) The point to point waypoint following is very naive and noone would ever use it, and he wouldn’t suggest anyone to use it that way.
(Carlos) What does waypoint following actually mean? What if we need to stop at a point, and then do something else?
(Matt) Mixing navigation with other tasks would be a mission execution.
(Carlos) What if we need different velocities at different section of the waypoints? Carlos has different implementation methods in the issue.Is P2P is the most basic primitive? Or should there be other primitives?
(Steve) He likes option # 1 better. We should push all of the complexity to application layer.
(Francisco) He thinks option 1 is the one that meets the requirements of what we really want to do.
Decision: Option 1 - PR #1338 is the best option, we will integrate into our repo
Next meeting: Dec 5th
Thanks to @Because for taking notes!