Work on a improved visualization of the tree. about behavior-tree HOT 13 CLOSED

For a free, open source and OS-independent visualization, OpenGL is probably the only decent choice we have right now. Do you have any alternatives?

from behavior-tree.

Indeed! I think it is a good thing to stick with OpenGL. What I really think should be improved is the graphics! So, it is a bit primitive yet... I mean... it is the first version and demands improvements.

Also, I think that the visualization implemented here could be an alternative if there is any unnecessary resource demand by the use of OpenGL.

from behavior-tree.

That is with graphviz, I have the same visualization on my python implementation of BTs (soon available).
That visualization does only outputs a png (or pdf) file. It is not an on screen visualization in which we can visualize which subtree is currently ticked (executed) as in OpenGL.

from behavior-tree.

Yes. That is right! The "static output" by graphviz is the major drawback. But, I must agree that it is simple and clean graph output. We should get a pretty OpenGL representation.

from behavior-tree.

I hacked some slight changes to the visualization on my machine - nothing major, mostly optimizing the spacing of the nodes at each depth layer - which made the visualization more useful to me. I'll post a screenshot later today.

from behavior-tree.

Awesome, keep us posted!

from behavior-tree.

Hey @int3ll3ct... That is awesome! Keep in touch.

from behavior-tree.

Here's my current slightly modded BT viz, note that I'm using the ROS-Behavior-Tree code as a baseline. My goal was to avoid nodes visually overlapping, which was a problem with my tree.

from behavior-tree.

@int3ll3ct Have you tried to use the PgUp/PgDown key to increase/decrease the spacing between nodes?

from behavior-tree.

Yes I did, but the existing viz algorithm overlaid my PanTiltServosOff and MemPriority nodes at any x spacing (the existing algorithm computed their centers to be dead-centered over each other). What I have done is no prettier, but will always maximize the space between nodes at every depth level. I also added a JSON parser for the BT visual editor at Behavior3.com, in order to create trees visually, and read in trees from JSON files, and I added more decorators and a couple of default actions. Out of the 3 BT implementations I was considering using, I felt yours dealt with the node "halt" concept the best, so I switched over to your implementation (I first started first using the almc implementation, which seems to be an ancestor of your implementation.)

from behavior-tree.

@int3ll3ct I agree, sometimes the overlapping can be tricky to solve wil PgUp/Pg Down. Fell free to upload the improvements.
I am in contact with the developer of Behavior3.com I also had in mind of doing a parsing from JSON files (I have also asked him to add the possibility to draw the BTs from top down). Unfortunately this is still in the TODO list.

Yes, I was working with almc, he is my colleague. We now work on different topics, so I have decided to have my own library.

I feel flattered in having people using BT++, I honestly did not expect to have many people using it. Fell free to comment or request improvements/additions.

from behavior-tree.

The version 1.2 of the ROS library has an improved visualization.

from behavior-tree.

done in version 1.3

from behavior-tree.