Has Dino AI been improved?

[Kyd]

On ‎9‎/‎25‎/‎2017 at 4:33 AM, ducttapefixeseverything said:

Yup. It started sometime around when they changed rex hit boxes so it would be more difficult to trap the dragon. 

Now dinos think they have to attack the rex's butthole to do it damage.

Same with mosas. Alphas and normal mosas go after the tamed mosa's butthole.

I do not know what it is with WC and buttholes. I understand that it can easily be construed as a weak point on any creature, but come on. 

Our tactics should not consist of "Tank with one tame's butthole while another tame attacks the wild creatures butthole." 

Bravo! LOL! Slow clap!The "butthole" maneuver...

[LouSpowells]

3 hours ago, Lewiatan said:

What was exactly "dramitcally" improved? Don't tell me proper pathfinding is unrealistic exceptation for AI. Pathfinding was my the most anticipated AI overhaul during Early Access. Developers are not limited by technology only their skills. FEAR has outstanding AI and the game is from 2005. Granted is way more smaller game, but ARK could have implemeted similar features from there.

Well that's marked shift from your original argument that dino AI hasn't improved since day one. There's no arguing with that assertion, it's simply incorrect...but apparently the subject needed changing.

To that end, I can only say this: Suggesting that Wildcard should utilize AI features similar to those found in FEAR and implement them into Ark is an assertion unbound by the constraints of prerequisite knowledge. I recommend learning more about modern video game AI implementation (particularly as it relates to server-hosted games with thousands of AI entities, hundreds of thousands of dynamic objects, and upwards of 100 clients) before continuing a discussion on the topic.
 

[waterKeeper]

26 minutes ago, LouSpowells said:

Well that's marked shift from your original argument that dino AI hasn't improved since day one. There's no arguing with that assertion, it's simply incorrect...but apparently the subject needed changing.

To that end, I can only say this: Suggesting that Wildcard should utilize AI features similar to those found in FEAR and implement them into Ark is an assertion unbound by the constraints of prerequisite knowledge. I recommend learning more about modern video game AI implementation (particularly as it relates to server-hosted games with thousands of AI entities, hundreds of thousands of dynamic objects, and upwards of 100 clients) before continuing a discussion on the topic.
 

I believe when players make complaints like this, Lou, what they're actually requesting is quality over quantity, something ark missed by light years. 

 

I loved that they added lots of new dinos, but at the same time i was expecting an overhaul as part of it, before release. It was...misleading, to say the least. Fooled by my own expectations i guess. Otherwise, I would take quality over quantity.  

 

Realistically, we are still pretty much at square one. My expectations were WAY UP HERE and so from my pov, they've only moved an inch from the starting line. The dinos wander in aimless circles. The map is more like a store with static inventory that refreshes occasionaly that we browse, rather than a wilderness full of dinosaurs. 

 

I know there's more to it running in the background, and i dont mean any offense, but on my end, I just see Pac man level behavior.

 

There's a reason this game is multiplayer and a reason I get bored if not on official pvp and a reason they had to add really technologically out of place auto turrets to the game because our army of dinosaurs can't defend themselves, lol.

[LouSpowells]

I'd normally suggest researching the topic, but if you look at Ark AI and you see Pac Man level behavior I don't know what to tell you. Maybe try glasses.

[Lewiatan]

On 27.09.2017 at 4:08 AM, LouSpowells said:

I'd normally suggest researching the topic, but if you look at Ark AI and you see Pac Man level behavior I don't know what to tell you. Maybe try glasses.

Why did I shift from my orginal argument? Because I wanted to know the reasoning of another player why he thinks AI was dramatically improved. 

ARK AI movement works in this pattern in a loop: If you hit obstacle, turn X degrees and move forward till you meet obstacle or change direction after certain time. Automatic mobile vacuumers work in our houses in such way.  Therefore dinosaurs walk aimlessly around the map in circles. They show a bit of intelligent behaviour when they attack you or when they're attacked by other critters, but even then they try to walk into rocks instead of going around it. A few dinosaurs that behave different due to their abillities does not make ARK AI good enough in general. 

You have low exceptations for AI in ARK if you believe it was improved massively. No one asked them to spawn 26000 npcs on the map. Lowering that number would beneficial for performance and saved resources could have been spent on better pathfinding at least. 

 

[LouSpowells]

3 hours ago, Lewiatan said:

Why did I shift from my orginal argument? Because I wanted to know the reasoning of another player why he thinks AI was dramatically improved. 

ARK AI movement works in this pattern in a loop: If you hit obstacle, turn X degrees and move forward till you meet obstacle or change direction after certain time. Automatic mobile vacuumers work in our houses in such way.  Therefore dinosaurs walk aimlessly around the map in circles. They show a bit of intelligent behaviour when they attack you or when they're attacked by other critters, but even then they try to walk into rocks instead of going around it. A few dinosaurs that behave different due to their abillities does not make ARK AI good enough in general. 

You have low exceptations for AI in ARK if you believe it was improved massively. No one asked them to spawn 26000 npcs on the map. Lowering that number would beneficial for performance and saved resources could have been spent on better pathfinding at least. 

 

So your rebuttal is "nuh uh!"? Have fun with that.

[PeterManize]

Don't get your hopes up for AI improvement!

Let's be honest. For WC is a waste of money and time... I don't know if there are mods that can improve in the creatures behaviors.
Even if there is... it will be highly limited, most likely buggy as well.

I noticed that they "made some improvements" in the patch notes. Honestly I don't see them.

 

One improvement, was making creatures that are getting torpor damage to avoid going into water sources (so that they won't drown while unconscious). 
By experience, I've lost many creatures after that update in the same way. They still run to the water and drown.

Another note i noticed is AI pathing, I did noticed a difference to be honest, the recent creatures added to the game don't get stuck has much has they did before! 
But if you tame the old creatures, or the creatures with lots of clipping issues (Triceratops? Rex? Legacy Creatures?) You still have these headaches. It also happens randomly with the newer creatures don't get me wrong! (Therizinosaurus is the worse in this department!) 

[LouSpowells]

To be honest, and I'm being sincere here, I'm still a bit confused at what people are actually requesting or whether they're capable of understanding what is currently implemented and what is technically feasible. I've seen people say, "All they do is walk around" many times at this point. Well...yeah.

What is it that people are expecting from dinosaurs? Should they be starting a school for the blind? I guess they could make Raptors capable of opening the door to your kitchen. People need to put into words specific improvements...but before doing so, those making the requests need to be at least familiar enough with the concept of how Ark AI functions in comparison to AI in other games.

Don't get me wrong, pathing around trees and rocks could be better, but when you're talking about individual non-fixed entities¹ in the thousands performing large scale random pathing operations² performing interactions around various dynamic objects³ it's a different ballgame...but that's only part of the complexity considering a portion of these dynamic objects are stationary⁴, and another significant portion are other moving entities executing their own random pathing AI simultaneously⁵. At some point you have to accept that there are limitations to work around, particularly when you recognize how much this is already asking of a server.

Think about the AI you encounter in every video game you've played and consider the creature AI and the extent of their possible paths. Most AI creatures in video games are in a location/building/area with maybe a handful of non-stationary dynamic objects they might need additional AI programming to interact with...everything else (walls, stairwells, couches, tables, boxes) is stationary which significantly reduces the amount of AI required to interact with the environment. Quite simply, there are less dynamic objects to interact with and due to this less complex AI is required for the entity to perform its intended purpose. It's AI on rails. When you remove the rails to implement RNG pathing AI paired with dynamic interaction AI that will change based on what is being interacted with, the AI programming gets significantly more complex.

More complex? Aside from the map terrain, a vast majority of objects an Ark creature may encounter may or not be there. If a tree isn't there, the creature can walk straight, but if a tree is there, it has to walk around the tree. This sounds relatively simple, but is significantly more complex to program considering the size of the map, the quantity of AI creatures, the quantity of dynamic objects, the sheer volume of variables that must be taken into account for any of these interactions. A terror bird running through the redwoods trying to catch you has to run by a lot of these dynamic objects to catch you...and at some point it's going to get tripped up, especially if the server is already struggling to keep up with 99 other players, their structures, and their tames. Again, is it perfect? No...but what's the target here?

What's the point of comparison Wildcard should be aiming for when it comes to creature AI? DayZ zombie AI? GTA5 pedestrian AI? Pacman ghost AI? Left 4 Dead 2 witch AI? Starcraft Zerg AI? Rust helicopter AI? At the end of the day you can compare just about anything, but are the comparisons relevant? Not particularly...not when the intent and result are two entirely different beasts.

So what are we trying to improve here? What is a specific example of something that could be improved, how could it be improved, and do you have any suggestions for implementation of said solution? If people could answer these questions and provide the answers to the developers in the form of a request, we'd be far more likely to see some sort of change or improvement...but until then threads like these will continue to be impotent gripe fests. Peter had a helpful suggestion about avoiding water when torpor is being increased...but in all of the various threads in various forums on this topic, I've seen maybe a handful of valuable recommendations with the rest being little more than, "Improve AI, Wildcard sucks." That just doesn't cut it.

Yes, there is plenty of room for improvement, but it seems like a lot of people have unrealistic expectations derived from incomplete data. 

_{1) Wild dinos utilizing entity-specific (type of dino) movement AI
2) AI pathing is RNG with specific restrictions that are followed
3) Any non-terrain objects (rocks, trees, player structures, players, etc...)
4) Stationary: Dynamic objects the server must track that are "removable/harvestable," but relatively non-moving (trees, bushes)
5) Any moving AI creature from - meganuras to titanosaurs - both wild and non-ridden tamed dinos}

 

 

[waterKeeper]

3 hours ago, LouSpowells said:

So what are we trying to improve here? What is a specific example of something that could be improved, how could it be improved, and do you have any suggestions for implementation of said solution? If people could answer these questions and provide the answers to the developers in the form of a request, we'd be far more likely to see some sort of change or improvement...but until then threads like these will continue to be impotent gripe fests. Peter had a helpful suggestion about avoiding water when torpor is being increased...but in all of the various threads in various forums on this topic, I've seen maybe a handful of valuable recommendations with the rest being little more than, "Improve AI, Wildcard sucks." That just doesn't cut it.

Yes, there is plenty of room for improvement, but it seems like a lot of people have unrealistic expectations derived from incomplete data. 

 

 

Morellatops is a passive creature, but with at least one other specimen around they become defensive. A pack of Morellatops - if provoked - attacks until all members but one are dead, the last survivor will try to retreat.

 

I would like to see more dynamic behavior like this. Hyaenodon have this too, but b4 their release, morellotops was the only dino that showed me some intelligence, this was during like 2 years of ea. This proves they can give dinos some awareness of their environment, and react to it accordingly, the devs could take this further, dinos should be able to pick their battles better. All herbivores could use some herding behavior, throw some wild babiez in the mix, make momma trike aggressive instead of passive. I mean simple little things like that add a lot of immersion.

 

Give all the dinos plenty of animations. Some animations can have new attacks associated with them. The devs job is to make sure the dino knows when it's best to use each one or to flee. Other animations could just be things like warning  calls/gestures, non combat animations that are directed at other entities in the game. Like why does theri walk around acting oblivious to you, then have too immediately jump to full aggro, add a warning call as an intermediate stage of aggro, so a player doesnt accidently back into one, they could hear it first. And still other animations could just be aethestic, like a mammal shaking it's coat after leaving water.

 

Okay, passive tames. They could do something to try and save themselves right before dying, this would be optional probably. For starters their "taking damage" wounded cry could get louder or more urgent as their hp falls. A flyer could make a last effort to flee. 

 

The devs mentioned trip wire alarms changing the state of a tames behavior. Did this ever get implemented, I honestly don't even know. But basically, go further with this. Allow us to set "if this" :  "then that" triggers and actions with our tames. Something like: "if a nearby passive dino dies" then " all other passives switch to aggressive for 10 mins." I mean let us put some brains into our offline dinos. Or make interesting traps with them. Tripwire opens doors, pack of aggressive dimos get unleashed.

 

And the legacy dinos need to be overhauled, they dumb down the game a lot because most of them are only different from one another in their stats. Like raptors/carno/saber/wolf/terror bird is like the same enemy, different model. Close my eyes and swing my pike and it doesnt matter they have different names. I waited 2 years for this and still don't see it coming.

 

It would have been nice to give ark a solid single player pve mode. I don't expect anything like this anymore. It's not the game it turned into. Ark is fun for what is, its just that players all had their own visions for where it could go during ea.

 

And Lou, players did get misled by the devs during EA regarding dino ai improvement. For example (paraphrasing from memory): "dinos will be more or less active at dusk or dawn, maybe sleeping dinos" turned into one dino, the megalosaurus, sleeping. I think even migrations were mentioned too. I know these weren't promises, I'm not mad or anything at it, wildcard just made it seem like they were going to simulate an ecosystem at one point  but instead it largely stayed the same.

 

 

[LouSpowells]

Thanks, WaterKeeper. What I like about those suggestions aside from the specificity is that they all seem to be suggestions that, within reason, are feasible considering the limitations of the mechanics of the game. I don't necessarily agree with all of them (carnivores having similar aggressive AI, for example), but they're specific suggestions with explanations of the reasoning and a description of the outcome. Any time a customer wishes to see a change in a product the company should listen, and this is a good way to get good suggestions heard.

Whether it's a valid mindset or not, I think we can all agree that starting a suggestion with "The employees of this company are idiots and they need to fix ______ now!" isn't necessarily an effective way for a point to be delivered. Too many suggestions made to ANY companies start this way, and it's only gotten worse since the advent of online public reviews that nurture the "Let's see who can insult them in the most creative way" mentality.
 

[ruxxsilver]

I feel like I have to trap everything I want to tame now that dinos always run from me, I end up chasing things across entire biomes ><