Tactical Shooter AI - Asset Store Pack

I don't have a release date planned, as the last 3 times I said "next week," it obviously didn't work out. I haven't put any thought into pricing yet.

As for progress on the actual asset, I scrapped the behavior swapping system since the last time as I didn't like it at all, and am completely rebuilding it.

 

If you want any tester, i will be here.
I'm building a CTF game so it would be good to test custom behaviours (catch the flag and it get back to base, focus fire against who is wearing the flag, etc..)

 

Well, luckily the focus fire is already implemented, and you can also set agents to guard certain targets. The dynamic objects would work well for the flag once an agent gets within range... however, I've yet to make an "advance in the general direction of a target" behavior. I'm not sure if I'll ship with that or not.

 

Bumping thread!

 

Done with the custom behaviours and behaviour swapping in general. All you have to add your own behaviours is put it on the gameobject, set the AI's behaviour to custom and choose whether this will override the idle or custom behaviours.

 

I am watching this thread with great interest. The asset store has some good toolkits for building AI but none that are complete out of the box or those that say they are complete are just very basic. IMHO, when you publish, I think you will have the best complete AI system on the asset store.

I am pleased to see from your previous post, customization will be possible. Do you know if it will be possible to build customization modules in a tool like Opsive's Behavior Designer? The behaviour tree diagrams make it easy to debug and I find the behavior style logic perfect for AI.

 

Well, each "customization module," as you called it is simply a class with methods that are called from the AI's 'base' script. I don't see why you couldn't simply modify the Initiate or EachFrame methods to do whatever you need to do in order to make Opsive's Behavior Designer work. As long you update the targetVector variable in the "customization module" in order to provide the AI a position to move to, it should work perfectly fine.

 

Bumping again

 

Alright, I think Paragon AI is feature complete! After trying fruitlessly to swap out mecanim animations at runtime, I ended up just making an editor tool that lets you add states without having to deal with mecanim itself. There's a little more set-up involved, but only a couple seconds worth.



On to documentation! Quick question: Which of the following is the most useful/important to you?

1. Code comments
2. Written documentation
3. Youtube videos

I plan on doing all 3, but it would be nice to know which I should put the most effort into.

 

utube videos and written documentation

 

+1 for Youtube.

I find that 80% of the effort required to write documentation just covers the basics which are usually intuitively understood. It is important to have the written documentation, but I think Youtube how-to video's can cover this off quickly and also puts everything into a context that is easily understood. I would put documentation at an important 2nd.

 

2.
3.
The 1. may have already been done during the development.

 

In addition to documentation, I'm also making some custom inspectors to make life easier:

 

Some of the documentation so far:



It's a harder/more tedious job than I thought, putting everything down on paper.

 

Your efforts are going to make a good result.. Personally, I hope You made an asset for RPG

 

First draft of the documentation. Did I miss anything?

<Broken Link>

 

Yep. You need a Welcome/Thank you page. You need a how to contact page. You need a UFPS Integration page.

 

Thanks; I'll start working on those!

 

You are the man!
When we can test it?
Are there any tutorials or webinars planned for this good product?

 

Release is... I don't know when, actually. Soon. I think.

I'm making a manual (which you can read above) and also some youtube videos.

 

New demo!: <Broken Link>

I've added some new agent/weapon types and also improved the behaviors as well as a few other aspects.

 

Trying to work out UFPS support at the moment. There's a bunch of systems which aren't playing well with Paragon AI.

 

i dont like ufps , its overdesigned, animating everything is a much better way to design an fps system in my opinion, instead of doing everything procedural. ive tested my fps controller and i got 200 fps then switched my weapons and controller to ufps and framerate dropped to 55 fps(this is on a turbo system ) but i do like your AI system , seams solid and useable. does it include asimple aiming IK system?

 

I use to use UFPS. I did not like working with it. Doing basic stuff was unnecessarily over complicated and it has less features than RFPS. After switching to RFPS I couldn't be happier.

 

yeah needing to set the muzzle flash for every state alone is just stupid, as an example. while u could just parent a particle system to the barrel of the gun.

 

It does automatically aim the weapon towards the target by rotating the spine bone, but you'll need to make an aiming animation for the upper body.

Anyways, the UFPS problems seem to have been solved by just swapping out UFPS bullets with Paragon AI bullets. Onto youtube videos!

 

Here's a UFPS demo! You can fight against the AI yourself. Note that all the cover here is dynamic- there are no pre-marked nodes. That means that there's the potential for the AI to behave even better than what's shown here:

<Broken Link>

 

Are you going to do multiplayer? I'm thinking mainly co-op MP where you have a specific number of co-op AI that other players can take the place of to join your team.

 

Um, the demo is impossible. The AI knows exactly where the player is at all times (even after you die) and they team up on you immediately. So, way too difficult to even test out.

 

I plan on doing it in the future, but the initial release will not have multiplayer support.

It is supposed to be difficult.

The AI knows where you are for a user-designated amount of time after spotting them. This is so that they can predict where a target will appear from if it disappears behind a wall. I didn't design the system to account for the teleportation that occurs when the UFPS player dies.

 

okay. But it's too difficult to see how well it works with UFPS for the demo. Just my 2 cents. take it or leave it

 

I think I may have played it so much that I mis-judged how hard it actually was because I was so used to it. I'll look into making an easier one.

 

If you lower the damage amount and fire rate on the AI, it will make it easier to test out.

 

Here's a newer demo with a simpler map and fewer, weaker enemies. It should be easier (I think!). Regardless, it probably is a better demonstration of the system.

<Broken Link>

 

Great demo, but it's a little bit annoyning that the AI always knows where i am

 

In the package, you can change how long it takes for the AI to lose track of you. I set that to a long time in this demo because I wanted there to be a persistent firefight, rather than a game of cat-and-mouse.

 

Will you be adding Melee as well? Would be nice to have some AI that attack via melee whether that's a soldier/person/civilian who is out of ammo or has no ranged weapon, or a zombie/monster/humanoid creature.

 

FYI - in the SkyCity demo, I hid in a corner in the parking area so I could take them out with the shotgun as they approached. Well, one of them got all up in my space and I ended up standing on top of them. You may want to set a minimum range so that if they are within that range, they don't use the same space the player is in. Also it would look really odd to have 5 AI standing in exactly the same spot. Another thought is multiple teams (so team A, team B, team C, team D) where maybe team A & B are friends versus team C & D, but team B is friends with the player and team A isn't.

 

Melee is already supported, although it is not in the demo as the Bootcamp soldier doesn't have any melee animations- and trust me, you don't want to see my attempts at making one.

As for the glitch... I'm not entirely sure why that occurs. They're supposed to stop before reaching the player, but for some reason AI's which are supposed to use cover but can't find any don't. You'll note the guy with the shotgun (green) will stop some distance away from you. (Or at least, he's supposed to!)

Multiple teams are already supported.

 

will this ai system wrk fine on mobile devices too ?

 

I'm not sure, never tested it on mobile. If I had to guess, I'd say it would work just fine.

 

if the code is well optimized by u then it will work just fine for sure , because i am going voxel style so no weight from mesh/texture side just scripts should b good n optimized . its been may b 12 hrs since i gotta know about paragon ai n i feel like i m waiting for ages for the release , well some hard wrk n great results u got there . cheers mate\

Edit : I forgot to ask 1 imp thing will this support unity 4.6 ?

 

I'm using Unity 5 to develop this. I'm not sure how backwards compatible this will be; I know for sure some of the editor windows won't.

 

oh ok then it wont wrk as it needs o b exported via unity 4 , uity 5 package wont wrk on unity 4 . i too want to use unity 5 but the lighting is fecked up just cant get theit look good as unity 4 :x oh well

 

Unfortunately, there are some things that just aren't possible with Unity 4, namely creating mecanim blend trees from an editor window.

 

Hi ! I just saw that shooter AI (an other asset) is getting abandoned and that it's owner redirects us to your solution.

Could you please make it easy to plug our own pathfinding ? I'm one of those crazy guys making a full procedural game, I can't use the standard navmesh. You could for example just make sure all the pathfinding related code is in one class (like a facade pattern). I don't personally care for fancy plugin systems, I just need well organised code that I can override myself, without having to hack into the whole code base !

 

I'll see what I can do. However, some features won't work without the default navmesh, in particular the dynamic cover system. You'll have to modify whatever navmesh generation you use to manually place cover nodes.

 

I intend to run without a navmesh too. My game will be using the Y axis a lot, with dynamic obstacles.

 

Well, as long as you can automatically place these nodes at corners and behind chest-high walls, it should be fine.

 

could you add something to the AI like "line of sight" so after so many seconds of losing you they will search an area based on "the last sighted place" instead of tracking you behind walls.