The Future of Unity 4 AI Path finding

News & General Discussion
1

Does anyone know if Unity Team is planning on upgrading their path finding system?

I’ve been looking at Aron Granberg on A* Pathfinding Project and it seems much more advanced than Unity pro version.

Also he is due to release an improved Local Avoidance in version 3.2, I’m probably going to end up purchasing this, but would like to know if UT is happy with their current system and if they have any plans to improve it.

If anyone wants to see a comparison click here

http://www.arongranberg.com/unity/a-pathfinding/free-vs-pro/

If anyone has any info from UT can you please post a link.

General opinions on unity path finding system would also be appreciated.

1: unity Pro Navmesh If you plan to upgrade to pro for this feature then don’t. Unity doesn’t seem to care if you have a hard time with this, they don’t even bother making a video tutorial.
Although limited they have made an addition in Unity 4 click here Navmesh Obstacle is nice to see this implementation. I’m going to test this with multi-object sizes. (tank, humans)
Price: Free with Unity Pro

2: xaitMap:
Website Link / Unity Forum Link / Official Forum Link
This looks impressive. The development speed is great, new features regularly.
Negative: I don’t think this system will do well with multi-size object navigation.
Price: $249 ON SALE

3: Aron Granberg pathfinding system:
Website Link: / Forum Link / Official Forum Link

  • Multithreaded + new local avoidance to be added soon and looks great, check his demo I have to test this with multi-object sizes.
    Big Positive : Multithreaded
    Price: Free / Full version $100.00

4: SimplePath:
Webiste Link / Forum Link
I haven’t checked it out yet, will add more info soon.
Price: $60.00

5: Dynamic Navigation by brilliantgames :
Forum Link
It’s amazing whats been created here in such a short time. 3 months fulltime and have a look how advanced it is. I think that brilliantgames will probably become the best in time, it’s rapid development and the direction he is going is the best way to go. From what brilliantgames has said in another post, is that he wants to create the best system with multi-object size system. Also implementing some other important features like: Ai being able to take cover ect, this is definitely the way to go, I mean you have this mesh using up ram so why not use that grid mesh to do other things. Smart move!! Please note this version is still in beta stage.
Price $50.00 Asset Store Link

6: PathFinder By Song-GameDev :
Website Link

2

Unity 4 added quite a few features over the 3.5 version (which was the first one, just to mention).

Arons is definitely much more advanced but his had soon 3 years to mature, Unitys solution has not gone live till half a year ago or so (don’t recall the exact 3.5 date anymore sorry) so for a fair comparision on ‘where they are going’ or if they move, you should potentially compare their state at the same time of existance.

For usage naturally, you might want to use Arons solution especially if you need the avoidance or multiple nav meshes etc

3

ahh Dreamora I was hoping a man with your wisdom would give an opinion.

I haven’t heard anything from unity mentioning anything about new added features in U4, is there a link i can look at?

“you should potentially compare their state at the same time of existence”

yes I agree with that way of assessing, but that would depend also on 1 person from unity is working on this vs 1 person “Aron Granberg” that way I could estimate the potential progress. maybe unity has 3 people working on it, by any chance do you know how many? hehehe that’s probably a silly question, secret insider info required.

Thanks for your input Dreamora

4

would require me to have the installers still around but from a list I’ve here:

and if I recall right more has been mentioned at Unite but might not needfully have made it in.

My glass ball unluckily has been in service for months now, but I’ve been told that its going to be fixed soon ™ :wink:

I’m not even sure if Emil ‘AngryAnt’ is related to it though I really hope so. Emil wrote Behave as well as Path, Arons original competitor in the Unity space basically, and I’ve some high hopes that Behave will become a full Unity feature (similar to my hope that stramit joining UT means that his node shader editor will become an official feature at some point in the future) but thats stuff that at best UT staff knows, if at all yet

5

My question is why didn’t Unity buy Aron’s project (and/or hire Aron) and integrate it into Unity core? It seems rather Microsoft-ish, where they want to build their own technologies from scratch and wind up way behind.

6

DallonF that’s a good question, maybe Aron didn’t want to be hired, who knows…
@Dreamora : thanks for the info + links. I checked out behave by AngryAnt. Behave looks promising but the tutorials are lacking. So I’m not going to going to bother looking into his path plugin. + slow development as he is very busy.

7

Yeah god forbid you’d compare the efforts of just one man part time to a million dollar company dragging it’s heels on it. I’m thinking Aaron put effort in.

Plus aaron’s really rocks in the way you can just set up a bunch of random gameobject targets as children of a game object, then hit bake, then get back a valid node-based path. Navmesh isn’t very useful if say you’re making a side view platformer or something else, but aaron’s one will work perfectly fine in cases like that.

8

So would it be possible with Aaron’s to bake a Navmesh per Prefab (tile) and combine them at runtime to make a dynamically generated map?

9

It’s a shame, i just went to purchase Aarons A* from his site but having problems with his shopping cart.

10

Dreamora doesn’t know anything… because he spends all his time posting on the forums :smile:

Aron Granberg’s is the best pathfinding system I’ve used. I bought pro and it’s very happy, and Aron gives good support.

Stay away from SimplePath, I had tons of runtime errors when instantiating agents and adding pathfinding to them, things just didn’t make sense.
When I posted a mini sample project with my problem and asked about it on the official forum thread, the author just ignored me, both times.

You can get it from asset store too… or just mail him.

11

Speaking of which, I really wish Unity would spend more time doing QA testing in approving assets before they are on the asset store. Many assets aren’t as advertised or are faulty with such bugs.

This seems especially needed since Assets provide many missing functionality in Unity (such as runtime pathfinding with changing terrain geometry).

@Meltdown - have you tried rainone?

12

I remember looking briefly at RainOne a while back, but never gave it a proper look.
I’d suggest using the free version of Aron’s library to see if it suits you needs, then go from there.

13

I tried RainOne a while ago but it wasn’t a positive experience, their documentation is very poor and the lack of debugging features and it’s closed source meant when things don’t work right out the box figuring out wasn’t working was a tough move. The only good thing I’d say about it is that their support respond really quickly to queries.

Aron’s library is superb and has some really good features coupled with a good set of documentation too – I’d recommend that to look further into.

14

Closed source is a no no when it comes to AI IMO

15

it’s amazing the vote results are 33 to 0.

hoping something’s done about it. otherwise, everything new in Unity is all just half-ass: “oh, we have path-finding AI-ish, and we also have animation-ish, and new particles-ish, we have mobile shadow-ish (only on certain non-tegra devices), we have android support-ish (where you have to write plugins for the other 50% of the missing functionality we forgot)”

16

Yea, Unity’s current solution is quite disappointing, I wanted to use it to test a new game type, and all I needed was what I thought was simple ground movement, and simple flying movement. Ultimately, it’s a well integrated, polished solution designed to make demos of one type of walking character.

17

I’d go with Aron’s pathfinding system. Even though I didn’t fully understand pathfinding, I was able to sub-class his Ai code, into my AI characters. Took me a week just write their FSM behaviors, but an hour to mode his code.

18

why are people surprised that everyone voted yes… sorry, its just a pointless poll, similar to ‘should unity fix their garbage collection’

19

I was wondering if anyone had an answer to this question, because it’s what I’d currently like to do. I would like to prefab a bunch of rooms, then have them combined at runtime by the user. Aaron’s library is looking the best option for nav, so this question caught my attention.

20

That poll…