Abort flag in Tasks at Execute Task service

Question

Abort flag in Tasks at Execute Task service

MFernandezCarmona opened this issue 5 years ago · comments

Manuel Fernandez-Carmona commented 5 years ago

Hi,
It seems that we cannot abort tasks on the vehicle execution node. We have an example code to call it https://github.com/OrebroUniversity/navigation_oru-release/blob/master/orunav_vehicle_execution/src/send_abort_task_main.cpp, but it seems that it's not handled.

Is this obsolete somehow? What would be the approach?
Thanks

Henrik Andreasson · Answer 1 · Wed Mar 11 2020 04:28:52 GMT+0800 (China Standard Time)

Hi, well... a lot of stuff has been in place working back in the days. Now to be honest I'm not really sure what is working and not there ;-). As there been many changes in the namespaces and this has broken some stuff in the past maybe that would be a place to start? Otherwise I would suggest that you add what you need in the code directly. Best, Henrik

…

________________________________ Från: Manuel Fernandez-Carmona <notifications@github.com> Skickat: den 9 mars 2020 13:02:42 Till: OrebroUniversity/navigation_oru-release Kopia: Subscribed Ämne: [OrebroUniversity/navigation_oru-release] Abort flag in Tasks at Execute Task service (#38) Hi, It seems that we cannot abort tasks on the vehicle execution node. We have an example code to call it https://github.com/OrebroUniversity/navigation_oru-release/blob/master/orunav_vehicle_execution/src/send_abort_task_main.cpp, but it seems that it's not handled. Is this obsolete somehow? What would be the approach? Thanks — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#38?email_source=notifications&email_token=ACUTFHU6CTJFCICVW6NB26LRGTLGFA5CNFSM4LEHGAU2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4ITRMG5Q>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ACUTFHS4THOLTU2ST6VNYB3RGTLGFANCNFSM4LEHGAUQ>.

Manuel Fernandez-Carmona · Answer 2 · Wed Mar 11 2020 17:26:42 GMT+0800 (China Standard Time)

I know, this is so big and complex that I'm always afraid of breaking things. I checked for the namespacing, but it as "abort" is inside "execute task", and "execute task" service works, it should too.
So I searched inside vehicle execution node code for any use of the word "abort" but no results. I think this may be something removed from the code before uploading it to github.
Could it be the case?
Thanks

Manuel Fernandez-Carmona · Answer 3 · Sat Mar 14 2020 01:30:15 GMT+0800 (China Standard Time)

I think I can work with update trajectories and avoid aborting. But in the future this may be relevant...

Manuel Fernandez-Carmona · Answer 4 · Wed Apr 22 2020 16:27:51 GMT+0800 (China Standard Time)

Hi,
I've been working on trajectory updates, but there is a limitation in this approach: "computeTrajectoryChunksCASE2" method in vehicle execution node needs at least 3 points in the new trajectory to work. As I'm calling coordinator to trigger this, I need to provide at least 5 points due to the extra processing time.
So, updating trajectories to force the robot to stop only allows me to do it with a huge delay. I.e. the robot will still advance almost 2 meters after the signal is sent.
I've tried to force the "brake" state implemented in the proximity areas, but from that I cannot reset the trajectory pool to create a faster "abort".
@h72a65 @FedericoPecora @ksatyaki you can see the problem in this linked video:
https://www.dropbox.com/s/o19izprgoa9o45r/Abort_as_task_update.ogv?dl=0

Martin Magnusson · Answer 5 · Fri May 08 2020 21:28:00 GMT+0800 (China Standard Time)

Just discussed this with @h72a65 and a better way to brake fast would be to compute and apply the new "stopping" trajectory segment inside the vehicle execution node, which would save the round-trip via the coordinator, if I understood correctly.

@MFernandezCarmona could you look into the vehicle execution node and see if you can apply the abort directly in there? And in case of problems, a direct email to henrik.andreasson@oru.se is more likely to be seen than a mention here.

Manuel Fernandez-Carmona · Answer 6 · Fri May 08 2020 23:35:37 GMT+0800 (China Standard Time)

Agree. One of my attempts was to use "sendBrakeCommand" inside a callback to force a stop and then send a new trajectory, but never managed to change the trajectory, because it complained about being on the wrong state.

https://github.com/MFernandezCarmona/navigation_oru-release/blob/master/orunav_vehicle_execution/src/orunav_vehicle_execution_node.cpp#L1511

Maybe that it's useful with some changes. I'll forward this to Henrik to see his opinion.

Manuel Fernandez-Carmona · Answer 7 · Wed Jul 01 2020 23:12:01 GMT+0800 (China Standard Time)

I have implemented an "abort" behaviour. It basically stops and clears out whatever is remaining from the ongoing trajectory. We can discuss it in detail at the pull request.
Cheers