When you use MoveTo BT node, with a target set as Actor and your AIController is DetourCrowdAIController, then the pawn will not follow the moving target.
How to reproduce:
Create new third person blueprint project.
Add a nav mesh that covers the level, default parameters (with 4.7.2 I also had to set update at runtime in project settings, looks like it´s not more necessary)
Put attached files in Blueprints folder.
Put the “Bot” actor in the level
Play and run around: the bot will follow the player character. This is the correct behavior.
Edit the Bot class, set Pawn Controller Class to DetourCrowdAIController
Play again and run around. Notice how bot will not update its target when player is moving.
I am testing out the repro steps that you have provided and got the same result. However, I was able to pinpoint the problem. The AI is getting caught on the “Wait” after movement, making it look like it will not update. I fixed this by adding a selector:
The difference now is that the AI must reach the Move To location before it can update again. I am now discussing the detours functions with some AI developers on Fortnite and will get back to you with more details about why it causes that.
Thank you. I had the “wait” in sequence just to have a moment to move away from the bot in the beginning. I tried different starting locations on the map to see if it´s not location dependent. For me it’s ok like you did it too.
I tested this internally and the DetourCrowd now does not cause the updates to be such a problem, however it does cause the AI to rotate oddly as the updates are not constant.
I’ve entered a bug report (UE-13275) for the difference in movement after I was able to get confirmation that they should behave the same.
I’ve entered a bug report (UE-13275) for the difference in movement after I was able to get confirmation that they should behave the same.
We use “controller preferred rotation” in bot movement and “allow strafe” in “move to” node. With that we noticed 2 problems with rotation: AI can walk sideways or backward and it can start turning left-right very rapidly. We solved that by creating a blendspace with forward-strafe-backward walk and by setting target weight interpolation speed parameter to 1.
Also, difference in rotation update can come from some parameters. I notice that rotation is not fluid in the example, but in our game it´s perfectly fluid. I cannot really find which parameter are causing the difference.
