RVO Obstacle push AI out of path

Beginner Questions
#1

AI want to move along path, but RVO Obstacle always push AI out of path. How to fix it.

#2

Hi

Where is the RVO obstacle in that video?

#3

red line is obstacle’s edge

#4

Huh. An RVO obstacle shouldn’t behave that way.
So you are using the RVONavmesh component? Does everything work if you disable that component?

#5

Disable RVONavMesh, Everything is ok! But AI will go out of navmesh when steering behavior calculate target position out of navmesh.

#6

Strange. Does anything change if you modify the ‘wall height’ field?

#7

I use default wall height (5m), I change 1/2/3/4, It is same result.

#8

What movement script are you using? That doesn’t look like one of the built-in ones.

#9

my movement code from LegacyAIPath class

image.png1440×873 62.4 KB

#10

Ok, could you draw in the scene view the velocity that your movement script tells the RVOController to use?

#11

#12

If it is the red line that is your script which is ordering the RVOController to move in that direction, that means it is not the RVO system that does something strange. Does your code interact with the RVOController in any other way?

Also, what are you doing with those gifs? They turn almost completely green sometimes:

#13

so sorry for bad gif! red line is calculated by rvo controller.

#14

So could you show me the one which is what your script tells the RVOController how to move?

#15

blue line : desired vector by movement ; red line: Calculated Vector by rvo

#16

Hm… Since you are using many legacy components, do you perhaps have a field called wallAvoidForce or something similar on your RVOController or any other component that you are using on your agent?

#17

Yeah ! But Obsolete! I am not use it.

#18

Hm… Strange. It looks almost exactly like there is some wall avoidance force being applied. I cannot replicate this in the latest version with updated components though.

#19

thank you so much ! I will try latest version …

#20

Hi! I update the latest version! Everything is ok. But I want to know that it must be recalculate path when setting target?

Is it possible that setting RVOAgent’s desired target without path finding directly?

  1. calculate AI’s desired direction in game context
  2. rvo controller set desired target
  3. AI will go into rvo obstacle

    I want to implement that AI will move along obstacle edge when collision with obstacle. Is it feasible?