search

robotology / react-control

Framework for reaching with whole-body obstacle avoidance.
GNU General Public License v2.0
0 stars 0 forks source link

Integration planner #13

Closed matejhof closed 7 years ago

matejhof commented 7 years ago

Mainly contains additions that allow additional targets to be specified for react-control. The reformulation is partially generic (for n targets), but in some parts, it is hardcoded only for the elbow. That is react-control can accept end-effector only target or end-effector and elbow. The targets can be obtained from rpc (only end-eff) or streamed from supervisor/planner.

Contains also corresponding additions for logging and matlab plotting as well as iCubGui visualization.

matejhof commented 7 years ago

How do we go about this here - @pattacini , @alecive , @towardthesea ? I think it is stable enough and we should merge it to avoid that we have a massive branch with a lot of stuff in it. The basic goal - multiple targets (more precisely - end-eff and elbow) tracking - has been accomplished. Also, @alecive and his student are planning to use the react-control soon in the Baxter, so it would be nice to consolidate it here I guess.
@pattacini can you perhaps add the tag before the merge in case we ever need to come back to the single-target-master?

pattacini commented 7 years ago

Hi @matejhof

@towardthesea is still testing this new branch on the robot. There are conditions that need deeper investigations indeed. Hence, I'd really wait before merging.

The history is not tangled, so there's no risk to mess up.

matejhof commented 7 years ago

Ok. Some of the problems under investigation (like IpOpt exit codes) are may not be specific to this branch though.

towardthesea commented 7 years ago

Hi, The testing on robot will mainly focus on considering why the globalTol has never been satisfied, in the case of reaching a position (with or without planner). The situation leads to conflict in robot behavior if there is another moving command (through Cartesian Controller, for example) sent to robot. Sending stop command to reactCtrl can be an option but it is not robust!

matejhof commented 7 years ago

I see. Go ahead then!

matejhof commented 7 years ago

Merged into master. Main functionality added:

Known issues: disableTorso on causes segfault Next: A new devel branch will be started.