Pet Pathing - The real problem (i have verified)

[shadowgod]

i have already let a programmer, yes a real to life AO programmer know about this, and he acklowedged and made comments about it to me last night. so problems are known to the devs.

This is what i have found to be the problem with pets in missions.

so, to give you some insight on how i study pathing in missions: i run missions 1st person with that little map open. i have the monsters, players, and machine updgrades to the map, so i can see my pets in it.

Running around in missions, i have noticed that if there is good distance between me and the pets (i.e. they're not directly behind me) they will just about 100% of the time hit the wall when i go through a door. Also, i'm starting to see more missions with "S" shaped corridors, these actually have helped me prove the problem i see with pathing.

are we ready for the reason? here we go..

Pets path indoors as though there are no walls or obstructions between it and its master or target

There are ways to then manipulate pets while in missions to path how you want them to. Example: if you go through a door and the pet hits the wall the door is on.. do not go back into the room the pet is in, rather, get at an angle that is line-of-sight through the door to the pet. the pet will then (by my testing) come to you in the new room.

those "S" shaped corridors, i found the only way to get a pet through them is to just terminate them and recast after you get out of it. my pets keep going back and forth in it with the ocasional stop to hit a wall. as though its confused trying to figure out how to get closer to its master with the walls in the way.

anyhow. thats my conclusions. will post anything that changes.

[__________Curmudgeon]

This wasn't known before?

Dang, and here I thought it was obvious. Heck, if you want to see it first person, try to autofollow a person through a mission. Then you too, can slam into walls.

When non-pet using people tell me they can't function with pets, I just tell them they need to learn how to move with a pet.

The easiest analogy for me is the difference between driving a little sports car and driving a big rig tractor/trailor. When you are by yourself, you can whip through doorways, and make turns on a dime, like you're driving a sports car.

But when you have a pet, you need to move like you are the tractor, and the pet is a big old long trailor that you are towing. If you go through a doorway, you need to keep going straight through that doorway until the trailor clears the doorjam, before turning.

Am I defending the pathing in this game? No, it's a pain in the butt. But once you get used to the fact that certain rooms are certain trouble, once you get used to always moving through doorways as if you have a ten foot stick of dynamite sticking outta yer butt behind you that you better not bang into anything, once the sound of your pet breathing is an essential subconsious go-signal, once you have the Map Reader upgrade that let's you see where your pets are.... it all starts to come together nicely. Usually.

[__________THEDEACON!]

Yep, this is exactly what I said. The collision detection in missions is fubared. Part of this, I believe is because of the "content on demand" that missions use. FPS games for example have static maps and the bots in those "nodes" (imagine connect the dots) to follow their way through the bots. At least this is how the "reaper bot" in the original quake mod was developed.

Imagine a set of legoes scattered on the floor. When you hit that accept mission button, the game quickly assembles those legoes into a map for you, which is why no two missions are exactly alike. The malleability of AO missions also make node based pathing nearly impossible.

So is there a fix? Maybe. Tim Sweeney of Unreal Tournament fame created a new type of pathing for Unreal Tournament that doesn't use nodes. It's been years since I read the interview and his developer notes, but I think he did it so that bots wouldn't experience trouble in user created maps (don't quot me on that).

To be honest, I never really looked at how it worked, but if I had to hazaard a guess, I'd say that each solid object is given it's own "bumper" that a bot could see and avoid. Bot detects that bumper and goes around the object.

So in AO, maybe each piece of the mission could be given this "bumper" code to inform the pet to go around? I guess that solves one problem, but telling pets not to run into walls isn't going to get pets through the doorway any easier.

The easiest solution to this (and it would add an incredible amount of realism to the way bots and pets work) is to make mission designs STATIC and to implement node based pathing.

Take BS missions for example. Maybe have three for four basic designs that are STATIC STATIC STATIC and when you implement the nodes, MOBs will instantly know how to get through doorways, through hallways, around tables, etc, because all they are doing is going from node to node (like a big string of connect the dots).

My wording sucks on all this, but anyone that's ever worked in game development probably knows what I'm talking about.

1) STATIC mission designs.

2) NODE based pathing.

3) problem is fixed.

[shadowgod]

well, apparently from what the dev mentioned was, the pets are not following the nodes (i believe he used that term too) that are already defined. which i'm sure each room of a missions is just a tiny lil `module` that are randomly thrown together. each room has their own defined path that pets should follow. but obviously, the pets do not follow them.

we might have know it, but it appears the devs did not know the full extent.. but now they.. at least one of em.. do

[ntrox1]

I think the node idea is complete enough in its theory, provided the nodes terminate in the doorways. If the gamecode is not recognising the nodes then thats the issue that needs fixing. Overnight we will have mission pathing issues resolved.

OK, what about outdoors? There are many locations that have invisible walls. WA by the booths has a hidden wall. HITW has a shedload of invisible walls. All of these walls stop the demons pathing. Not sure about balls though.

[__________THEDEACON!]

Originally posted by Ntrox The MP Gimp
I think the node idea is complete enough in its theory, provided the nodes terminate in the doorways. If the gamecode is not recognising the nodes then thats the issue that needs fixing. Overnight we will have mission pathing issues resolved.

OK, what about outdoors? There are many locations that have invisible walls. WA by the booths has a hidden wall. HITW has a shedload of invisible walls. All of these walls stop the demons pathing. Not sure about balls though.

Both of my balls stick alot too

[Mysticknight]

this still doesnt explain the pathing problem outdoors

tell a pet to attack a mob and it just sits there
nothing in between the pet and the mob
no trees rocks or buildings just wide open space

[Xtremtech]

Lol! I certainly hope they know this. This and other suggestions about the symptoms and possible causes were included in the pet pathing report which Cosmik said had received a good response from the dev team months ago.

Still it's certainly excellent that another dev is now in the loop. Let's hope that the situation is as clear as getting the pets to follow their nodes properly.

Having said all of this, we should be careful not to oversimplify the problem.

I run through solo and team missions all the time. I usually have two meatball pets following me: Trans Enmity and Bela. Both are following me at the same time, through the same rooms, in the same conditions... yet the enmity is the one that gets stuck most and the Bela is extremely well behaved (for me ).

If it's simply that the pets are not following nodes - then why is it so much worse with the attack pets?

If it's as simple as this, why do pets not go uphill or over fences or sticks on the ground when outdoors....? Note: we can run over most of these obstacles without jumping or changing course... why can't our pets? It's not a collision thing per se... since many of these objects would appear to be collision free graphical objects.

If it's as simple as this, why is it that pets still occasionally run in the opposite direction from their target for a great distance before turning round and coming back to finally reach the target... or more often they run a long way away and get stuck coming back along the same path they just ran...?

My worry would be that if we encourage the devs into thinking that simply solving the room node problems will solve pet pathing issues - then we'll have patch after patch full of mission design changes and no root changes to the way pathing works or addition of /pet warping commands.

We've already seen that while mission design and improved following of nodes does improve things with pet-pathing, it is unlikely to provide us with a substantive change that will satisfy all. Additionally, every time a new environment is added or an environment is changed there is a reasonable risk that that change will mess up pathing again - so node layout etc is not necessarily a long term consistent solution.

Let's continue to lobby strongly for substantive change and some method of retrieving pets manually when they do get stuck.

X

[Xtremtech]

TheDeacon,

Static mission room design would be pretty cool for pathing but of course slightly less cool in the experience overall.

The 'bumper' collision detection mode of pet pathing can be horrendously resource hungry and would like as not need to be implemented on the client side - causing CPU resources issues for the player which result in lag. At each 'step' the pet works out where you are, if something is in the way and the quickest direction to move to get to you. If you do the calculation on the server then you see more network latency lag in the pathing.

Those very lag/CPU load issues might in turn result in "hesitation pathing". For example, the pet starts to work its way around an object in your general direction. Due to lag though you are already somewhere else. By the time it realises that you're somewhere else and turns to start going in that direction, you're already somewhere else etc etc.

With node pathing you have the benefit when it works of predictive pathing. Ever noticed that when you run into a room towards a door and stop dead the pets often continue into the room through the doorway ahead and then come back to you once they realised you stopped? That predictive pathing means that most of the time - if they're following the nodes correctly - they'll keep up with you far better than having to always check where you are and follow you. If you had run through the door, they'd have already 'logically' been there....

I'm not up to date on the latest of the latest in pathing either... but it's always been one of those perenially difficult issues to solve in 3D environments. I even remember some pathing issues with dragons in UO, which was really unforgivable in an isometric environment....

If you want to try out for yourself how difficult it is for game AI to path when things go faster than you make decisions - go get all the very best run buffs so your speed is twice what you're used to and see how many times your superfast human AI runs into doorways/tables/bookshelves etc before it realises it's on the wrong path. That's what it's like for the AI.

X

[shadowgod]

yes, i too have noticed attack pet's pathing is worse than heal's in missions. most likely is due to their different routines. heal pet is designed to just stay with its target and not move around or need to defend, etc.

as for it flying clear across a mission then back again. this i'm starting to feel is master-path-predition meets lag. i know if i go up to a door sometimes, and it opens.. my pet keeps going on inside the room even though i had stopped. there may be a condition of excessive lag somewhere in there, where the pet thinks its master when several rooms in that direction.

i should really do more outdoor hunting to study exactly what pets do in this situation... will report when i find out anything i notice that hasnt been mentioned.

[Toxinator]

I dont care what the problem is

I dont care how they fix it

I just want a /pet warp till they figure out what the hell to do.

[Xtremtech]

yes, i too have noticed attack pet's pathing is worse than heal's in missions. most likely is due to their different routines. heal pet is designed to just stay with its target and not move around or need to defend, etc.

Yup precisely this is what we fed back in the Pet Pathing report several months ago. The major difference between attack pet and heal pet is its responses to aggro and it's reactions to the team being aggroed even when you are not. The heal moves around to heal other players almost as much as the attack pet moves around to attack mobs... so the movement pattern isn't different. Similarly, the nodes and paths that the pets take should not be different in any way.

Our su****ion in the report was that this is all tied to a change in the mob AI which was designed to allow mobs to support other mobs nearby and from several rooms away. As a result, the attack pet also went into combat mode when the team was aggrod or in combat particularly when its owner was above 50% agg/def. In faster high level teams, the pet is often trying to get to a mob that is aggroing a team mate a room or two away. It's no longer trying to follow you at all but instead is trying to defend the team by getting to the aggroing mob.

This is suspected as being why you'll notice a marked difference between pathing in solo missions and pathing in team missions. This difference has been reported for many months too. Many players who focussed on soloing found that the pathing was reasonable if not good. Players focussed on team play had considerably more problems.

You'll notice too, that when a team is in a mission all 3 mobs will come into the entrance from the main room almost every time - no matter how carefully you try to pull. In a solo mission it's possible to pull singles from the main room easily. The way in which you are being aggroed and the way the mobs AI makes them support each other is significantly different in soloing and team missioning. At the same time pet pathing is better when the mob AI is less aggressive and less likely to aggro from a distance.

After the mob AI patch, our attack pet started running several rooms away from the team to attack mobs who we hadn't even seen yet. The reason was that the mobs were in fact aggroing from rooms away and our attack pet, with a heightened support for the team that the mobs had for each other, rushed off to defend unbidden.

It was at this time of mob AI change that the big problems began to show with pet pathing. Previously, pet pathing had been an issue but for most players it had been a workable issue. After the new mob AI was introduced things went severely downhill.

It's clear that mob AI and pet AI are shared code. The attack and defence patterns that were changed in the mobs also affected attack pet AI in particular. We also saw clear signs in the "rubber banding" of attack pet and mob that used to happen a lot more back then. Both would run up and down an imaginary line back and forth in opposite directions.

It is also seen in the doorway behaviour where a mob will be on one side of the wall next to a door and the attack pet on the other. If you watch carefully, you'll notice that sometimes when the mob warps through or runs through the door, the attack pet perversely decides to go round too at the same time and they are still on opposite sides of the door....

The heal pets first became considerably better in a release where the heal pet no longer tried to attack in some instead of healing (as it had before in some situations). This change was stated in the patch notes at the time. They removed the aggro parts of the AI and the heal pet started pathing excellently.

Now these changes may just have coincidentally come at the times when improvements and worsening was noticed... but it could also be strongly linked.

If the node pathing were the only thing at fault then mobs, attack pets and heal pets should exhibit the same behaviour whereas in reality we see that the Pathing AI fails mostly when there is aggro patterns involved.

It's still not a complete answer to all of the issues grouped under the "pet pathing" headline.... but it seems to be a key feature. In the pet pathing report we made a request that pet AI and mob AI be separated. It's unclear whether this happened or whether the devs looked at this suggestion and decided that it would have no effect on the issues at hand. If so, then the above is largely coincidental. If not, then we hoped in the report that it would point them toward possible solutions.

As Tritoch says: However you look at the problems it's clear that since pet pathing has been a number one dev team priority for over 4 months and several patches now - it's clear that there are going to be no easy and fast solutions to the pet pathing issue. As a result, it becomes more and more important that pet owners get a manual way of bringing their pet to them from anywhere on the playfield when the pet is stuck.

One of the issues with this is that some players feel that the pet owner should not be able to do this while the pet is in combat - PvP issues are often cited - though others such as Zylina seem to feel that the pet being in combat should have no bearing on PvP directly. Unfortunately, in team missions the attack pet AI acts to defend the team - so the attack pet may be in combat even if you are not. Asking the team to stop fighting so you could retrieve your pet might not be a good solution either....

For a period we saw a behaviour where you'd command a pet to attack and it would get stuck. Immediately when the mob died the pet would warp to your side and growl (which was a little annoying as by then it wasn't needed any more). I haven't noticed this since coming back but have been hunting with the enmity with which it';s more difficult to tell and less noticeable than the demon. Is anyone still seeing this behaviour? I may do some testing.

Either way it's clear that a "warp when not in combat" piece of code is or was in the attack pet AI somewhere but that having it in there didn't necessarily help.

Many just want a simple manual warp under all circumstances.

That brings you a little up to date on some of the background ideas and thoughts on the pathing issue.

X

[Xephydes]

Originally posted by TheDeacon/Xcelsius


Both of my balls stick alot too

Now that's a sweaty remark!

[Sallust]

Pathing got a whole lot worse in 14.6. I wish you could have seen the change. Pets used to path semi-reliably outdoors at least. Now they treat most changes in elevation (no matter how small) as a wall almost. Only if they're stuck there for several moments will they find a way over sometimes. Usually they just weave like crazy.

I don't see any nodes being used outdoors for sure. Frustrating to no end in some cases.

[Shenu]

From what I can see, the easiest way to solve some of the pathing is to have separate pathing routines for indoors and outdoors.

Indoors the node thing sounds like a feasible solution. As missions is built from 'building block' rooms, if a new room gets screwed up nodes, there's just that one room to fix.

Outdoors, the rubberband or /follow type of pathing with sidesteps should do most of the job. At least for attacking. Most outdoor obstacles are small in nature, and the rest we can heard the pets past.

I had my attack pets do the 'stare dumbly' when i tell them to attack something in plain view, i also had it go straight in from the maximun distance i could keep the mob targeted. Something fishy is going on that i cant figure out.
Keep It Simple

Maybe a few different, smaller pathing routines for different 'modes' of behaviour would ease the server load, and improve pathing at the same time.

When not in combat, let attack pet use healpets routine, with an added check for anything to aggro on, then if it aggros, switch to the aggro routine.

One of the hearding problems in missions is that the pet stops at a distance from the owner, so you cant 'pull' it around obstacles by leading it by the hand.
Pathing / hearding wise, it would be best to set the distance to zero, but that would make the engis blind when they end up standing inside the bot. A smaller distance other than zero would help tho.

I do feel for the engis that dont have the opportunity to /pet terminate, as the nanoskill buffs to recast most often is not there.

Team mission pathing seem a lot better than a few patches ago tho, if i wait for the pets at every turn, i usually can make them come along. That is everytime i switch direction behind a door I wait for the pets to come through door before i start moving in the new direction.
Not the fastest way around the mission, but it beats loosing the pets.
Someone seems to be working on it, enough that it shows, and thats a good thing.

I cant see why pet commands need LOS at all, MP's are supposed to be psycics of sort, arent we?, and if an engi can make a bot that size, he surely can make a radio?? Dunno bout the crats there, but they'd probably send an email through the grid for a slightly slower response (as if anyone has heard of a fast working bureaucrat)

[zophee]

last time I teamed pretty often with fixers. I noticed that a large runspeed buff enables the pets to follow more reliably.
If someone can approve this, I suggest we should get a kind of runspeed nano for our pets so they can follow us a little better.
This especially helps in the S-type rooms.

For the node based pathing discussion:
Simply force the pets to use the 'door' node under all circumstances if moving between rooms; this should help alot on the corner-type rooms.

[Xtremtech]

Lol - nope quite the opposite is my understanding. Most people find that high speed missions cause even more problems.

What you will see with a moderate runspeed buff is that you run more 'cleanly' through the mission allowing the pets predictive node pathing to work a little better.... but with a large runbuff the pet simply doesn't keep up well with your changes of direction and beginst to get stuck more.

You could suggest that the pet speed be fixed at an optimal speed for its own movement. Unfortunately, the pet-speed must be the same as the owner's runspeed or the situation gets worse when in fast teams - because you're always having to ask the team to slow down so the pets can catch up...

It doesn't look like a pet runspeed change will be acceptable for the MP community on the whole - though the basic idea is a good one.

X

[zophee]

Thinking a little more about it seems I need to reformulate my understandings a bit ...

I see 2 different things:

1) pet is not near the master -> needs to path back to the master
In this case runspeed definately helps the pet to get back. I see the runspeed as another workaround like the warp- ability.
(btw: You can see this if you run away from a stuck pet - if you are some rooms away it will start pathing to your location, which I didn't notice at lower runspeed)

2) pet is near the master -> predictive pathing kicks in
This time things are as you mentioned XtremTech, but to me this seems to be a similar problem like getting stuck w/o runspeed buffs.
In this case I see the pets getting stuck most likely in corners, at furniture and beneath doors. This may be a result of the fact that pets alwas keep a certain distance to their master. This can imho only be solved in the pathing code. Perhaps pets could remember the doors their masters run through and pass these coordinates under all circumstances ...

finally I cannot the the problem what would be the bad thing if pet's runspeed is above the casters runspeed - perhaps you can enlighten me

[Xtremtech]

Sorry Darkzone, no it was me that misunderstood.

You're suggesting a pet runspeed buff that we can cast on it after casting the pet and so make it considerably faster than us. Which is a good idea.

It's advice I've been giving to people for some time in that you can use the Quickness buff on your pets to improve their performance a little bit. At higher levels and higher runspeeds though the effect is much less of course - as Quickness only increases runspeed by 30.

Having an extra line of pet-buffs that would increase the pet's runspeed by 50, 90, 140 and stacked with quickness might in fact be a very good idea indeed.

X

[ntrox1]

Your pet has the same runspeed as you, when buffed or not. If you have a very fast runspeed and leave the pet behind, it warps to you regularly. Its not an improvement in pathing, nor the pet being faster therefore being able to navigate quicker. Basically, whaat you don't see, don't happen. The pet will warp rather than work its way through rooms.

A runspeed buff for the pet will not help at all.