Progress of new Local Avoidance?

So sad to hear about the RVO license issue. Also I knew the new VO is under developing, can I ask the progress?


The new local avoidance is in the project as of version 3.5.
It is only in the pro version however.

Also note that in the current version, the lightweight example scene does not work as intended, there is a slight bug the scripts for that example scene, but the rest of the system should be working.


Thanks for replay! I think the new VO is not quanlified for production use yet. For now my project rely much on this plug-in, the lost of VO halt it, I tried UnitySteer, but not very good, so I would like to know current progress or estimated release time, if it’s too long, I will consider researching detour myself (which will be very painful to a junior like me).

By the way, if the new VO based on detour, will it solve the concave obstacle and narrow corridor situation like this:

I think the blog video you linked to used pathfinding as well as local avoidance.
Using RVOs, it is not possible for them to avoid a concave obstacle (possibly with some very specific settings for that specific case, but not in general), that’s why it is called “local” avoidance, it finds the locally best result.

The current system does not support obstacles, which is a drawback, however just agent to agent local avoidance is almost as good as the previous system.
Is it the obstacles you are referring to when you say it is not production ready or do you mean something else?

Yes, my "not production ready"only refer to the unfinished obstacle, also it’s the progress of which I’m eager to know about.

And about my link, I know RVO2 is unable to handle the concave scene, but Mikko’s “local grid pathfinding” seems solved it, so are you willing to add his solution to you APathfinding and beat RVO2 in the future?

Here is his latest research video, it looks not bad, just concern its performance(~1k agents?):

is it possible to use RVO controller on mobile?
I have a 3d scene, pretty light with 10 to 20 maximum agents and i use recast graph.

The local grid pathfinding is interesting, but I don’t think it is a solution I will add in the short term.
And if I did implement it, it would definitely not be able to handle 1k agents (I doubt his solution did either).

Yes, that is possible. 10 to 20 agents should be no problem.


The issue with obstacles is a blocker for me too. basically my problem is that it throws agents off the recast graph, since it doesn’t recognize the walls. Is there a way around that? or do I have to wait for obstacles?