Migration to use behaviortree_cpp

[ana-GT]

Hi, I was wondering if there are any plans in the near future to switch the dependency of ros2_planning_system from behaviortree_cpp_v3 to behaviortree_cpp. Thanks!

[fmrico]

Hi @ana-GT

Yes, we had a short discussion about this in #130 (comment)

The new version of Behavior Trees offers some exciting features that I want to include in PlanSys2. I had no time to work on this, but it is definitely in my plans. If you want to contribute in this direction, even with a direct migration, I will be happy to help you.

Best
Francisco

[ana-GT]

Hi Francisco,

The migration looks straightforward, unless I am missing something. I forked and created a branch named bt_v4 with my changes: master...ana-GT:ros2_planning_system:bt_v4 . Changes can be resumed as:

1. Replace behaviortree_cpp_v3 with behaviortree_cpp
2. Replace NodeConfiguration with NodeConfig.
3. Replace tickRoot() with tickOnce().
4. Add BTCPP_format="4" wherever appropriate (in tree's xml files and in code that create trees on the fly).
5. Replace success/failure _threshold with success/failure _count for the Parallel nodes that are created during BT construction.

The changes above were exclusively for migration purposes. I also added what I think is a fix for the is_valid_domain function in the popf_plan_solver. In its current form, if the user sends a valid domain and an empty problem, the function returns false, where I think it should return true instead. A unit test was failing for this reason.

Tested with the unit tests and with an example of my own that has 2 robots running in parallel, so I think this is working fine. I am happy to send a PR if that is appropriate at this time (wasn't sure given that the current ROS2/Nav2 is dependent on bt_cpp_v3).

Best,

Ana

[sarcasticnature]

I would have some slight concerns about migrating quite yet, as according to Davide's ROSCon talk v4 is currently in alpha/beta form. That means that there is still the potential for significant (possibly breaking) changes that could make early adoption a bit of a pain

While I agree that there are some exciting features, I'm not sure that the time is right as of yet (but that is a personal opinion).

[sarcasticnature]

This issue faces many ROS projects, which also need to consider when to make the switch. While looking through some of the changes, I stumbled across two relevant threads in Nav2 and BT.CPP itself

• Request: Behavior tree migration of 4.0 in ROS2 Humble. BehaviorTree/BehaviorTree.CPP#481
• Information regarding the nav2 stack upgrading to behavior tree cpp 4.0 and later versions ros-navigation/navigation2#3316

Personally I agree with the idea to hold off on switching until Iron is released, @fmrico what are your thoughts?

[fmrico]

Hi!!

My opinion:

1. Let's wait for Iron
2. Upgrade to 4.0 only if we take advance of the new functionalities. This requires a more extended discussion of what is new and how to use it.

best

[facontidavide]

Hi.

My plan is to release version 4.1 in Rolling and Humble the next week.
I will also coordinate with Nav2 to make sure that they use version 4 in Iron.

If you need assistance to migrate, please let me know.

[ana-GT]

Great news! Thanks for the update. Will keep an eye in Nav2 (and possibly PlanSys2 :) ).

[robodrome]

I made a PR upgrading from BT.CPP v3 to v4. #293 (comment)