I’m trying to create a new AI in Unreal version 5.1.1.
I did some research to decide what to use and learned in 2015 posts that pawn sensing will be removed soon, we should use ai perception.
However, I see that there is still pawn sensing in the current 5.1.1 version and it is still used in tutorials. What I’m wondering is, is pawn sensing updated, can it be used or should I use AI perception?
Hi, as far as I know pawn sensing is not being developed anymore for some years now (from this post here by mieszko Pawn Sensing vs AI Perception - #3 by Answers.Archive). I never used pawn sensing, only AI perception, so I cannot really tell you which one to use, but since AI perception is the new system and still gets updates I would rather choose that.
However, I see that there is still pawn sensing in the current 5.1.1 version
As long as keeping pawn sensing in the engine does not cause any problems, I don’t really see any reason for it being removed, since some projects may use it and could then not update to a newer engine version.
To get an overview about AI in unreal engine you could watch this video here https://www.youtube.com/watch?v=iY1jnFvHgbE
Yes, probably pawn sensing is an old and expensive method.
But it is used for experimenting with ai, if not for a finished game.
However, even in the tutorials shared in 2023, pawn sensing is still used. This makes me think that pawn sensing is the correct and used method. I’m sure there are many people who think that pawn sensing is the new one.
Like I said, this needs to be explained better.
I just discovered Pawn Sensing (I started learning UE from V5) and I will be using it because AIPerception is not consistent at all whereas my initial tests with Pawn Sensing seem to give a more efficient and instant sensing results.
AIPerception can be moody or random, sometimes AI sense based on aprox entered distances, sometimes not. And it seems heavier on Performance vs Pawn Sensing
I hope Pawn Sensing will be kept in 5.4 and beyond as I am still in 5.3
Can I use this thread as the most popular topic in this issue? - many will probably be interested in the transfer of events from AI_System to BP_pawn. This becomes relevant, because both Pawn Sense and AIPerception contain events (for example, detections) and transfer events to BP. Are there any standard mechanisms for managing behavior based on Behavor Tree, because it is BT that determines the possible actions of the pawn. Or am I misunderstanding something? - BT and AI are one control system, and (for example) Pawn Sense + BP (for logic) are another.
I found the following answer:
After a few days I found answers to my misconceptions about working with Behavor Trees. Let those who seek quench their thirst:
Common issues with Behavior Trees and things you should competely avoid. | Epic Developer Community
As for PawnSensing vs AIPerseption - for a beginner the second one will be preferable, because many things, including the debugger, are much more important and easier to set up at the initial stage of learning. As an example, you do not need to worry about processing events of loss of visual control of the bot through, for example, Conecheck, because PerceptionUpdate will do it for you. The setup is a little more complicated, but at the end you get a ready-made AI with literally two events of 2-3 nodes in each.