Local Avoidance pt 3

So it’s time for local avoidance again.
The A* Pathfinding Project (pro version) has included a beta version of local avoidance for some time. I decided it was time to do something about that “beta” label.
There is a great library out there called the RVO2 library. It has been used in both UDK and in Unity. Unfortunately it has a rather limited license. Not for commercial use unless explicit permission is granted, which is a problem since they do not seem to answer any emails I send them.
So I decided to take inspiration from the rvo2 library and rewrite it in C#. Which I have now done. I have also added new features, for example agents and obstacles can now have different Y coordinates as well as XZ coordinates, so agents at different floors of a building can navigate properly without “colliding” with agents at other floors.
I have written a script to convert navmeshes to RVO obstacles, so agents will properly avoid navmesh edges. I will later write a similar one for Grid Graphs and I hope to get graph updates to work fine with it as well.
Time to stop talking and link to a demo:

It has really good performance. On my laptop I can run 500 agents when I just make sure they don’t all try to go through the same point, and on my stationary computer I can run more than 500 agents with a good framerate. Currently the rvo simulation is set to a desired fps of 20 and everything in between is interpolated.


  1. George says:

    Dear Aron,

    I am very happy you will release updated version of the local avoidance :). As at current moment its impossible to create strategy games in unity3d with multiple units, as they stuck and cannot avoid each other.

    Waiting impatiently! Thanks!

Comments are closed.