everything is so damn hardcoded in to use strictly navmesh agent to point its just... ridiculously annoying. fuck it, gotta cut my time losses and do my own AI implementation. Trying to clean up behaviour designer from all the hardcoded useless shit is just incredibly frusterating.
Getting only Axis on where to move.
- Thread starter DageLV
- Start date
Deleted all behaviour tree shit and few hours later installed it again. Fk.
So to cache basic stuff I made _main_task :action class.
Then wander:_main_task.
I can't wrap my head around how do I make it stop wandering and do anything else.
It wanders. Okay. The ai controller (not part of behaviour designer) sees a target. Now what? How do I make it stop wandering and start the task to go after the target?
Every other visual scripting thing such as shader graph etc has input and output. I have no idea what is what in behaviour tree, there is only output, no inputs, no condition nothing
So to cache basic stuff I made _main_task :action class.
Then wander:_main_task.
I can't wrap my head around how do I make it stop wandering and do anything else.
It wanders. Okay. The ai controller (not part of behaviour designer) sees a target. Now what? How do I make it stop wandering and start the task to go after the target?
Every other visual scripting thing such as shader graph etc has input and output. I have no idea what is what in behaviour tree, there is only output, no inputs, no condition nothing
All three add-ons require an underlying navigation implementation. The proper approach is to implement the base classes with your own navigation implementation. This is what is included for the NavMeshAgent, and is what I did for the A* integration.
Instead of demo do you mean NavMeshAgent? The demo isn't anything special and as long as you are using a NavMeshAgent you can use any of the built in tasks. I mentioned above but there's also an A* integration. You will need to take the approach of implementing a new navigation system.
The built-in demo scene uses a NavMeshAgent but there are integrations with the A* Pathfinding Project and PolyNav for the Movement Pack. Behavior Designer itself doesn't use the NavMeshAgent at all.
Have you watched the set of behavior tree basics videos? You can follow the setup from those videos and expand from there:
You can use conditional aborts:
No. In demo you use tactical agent and it's deeply tied into tactical pack and also damagavle object which kinda means... They can be attacked or be attackers, not both?
I don't use a* as I said. And this has nothing to do with how the ai moves, but WHERE it moves. The final destination. Tactical and movement packs use agent everywhere, to do anything. Highly dependant on forcing people to use navmesh agent.
No haven't watched. I looked at what's in demos
There are interfaces that you can implement so it's flexible for attacking and taking damage. There's no reason why you can't design the tree to both attack and defend - the demo scene just shows a behavior tree with minimal tasks.
That's the thing - they don't require the NavMeshAgent. The A* and PolyNav integrations are a good example of not depending on the NavMeshAgent.
So there is a* for tactical pack and movement?
And the demo. One has tactical. Agent component which is deeply embedded in whole tactical pack and the damagable component, which is used for detecting targets etc.., didn't look deep how far is that embedded, but tactical Agent is very deeply embedded into everything,
Last edited:
Yes, it's on the downloads page with your tactical or movement invoice number: opsive.com/downloads/
TacticalAgent is, but TacticalAgent is just a generic class structure that is not tied to any navigation system. TacticalAgent is an abstract base class that contains references to the interfaces IAttackAgent and IDamageable which makes it extendable. The NavMeshAgent and A* classes implement TacticalAgent with their own navigation implementation. This is the same thing that you would do with a new navigation system.
oke found it. Movement pack has polynav support, isn't that just unity navmesh? Tactical doesn't have it. There are shit ton of addons for behaviour tree. So the tactical, movement and formations just has implementations, not addons per say?
Gonna try to make it work now tho.
its still very embedded with a*. Do you plan to release no-integration version?
Cause as of now, its very annoying if i have my own pathfinding, its very annoying if i have my own vision, its very annoying if i have my own shooting method. Now its mandatory i use some useless attack struct, some damagable struct and some more useless garbage which my vehicle does not need at all, but because its very deeply integrated in behaviour trees, i have to use them. its just... fk... extremely annoying to be forced to use someones else's navigation system even if their navigation system works nothing alike what you want, such as navmesh agent or a* agent.
Just a version with no excess garbage. For demos, fk it, a basic script which moves to Vector3 destination, looks at vector 3 and shoots at Vector3. Easy to get rid of and replace with my own, unlike what it is currently.
im not using some fucking first person controller, no fucking asset store pathfinding projects, no fucking astars or navmesh agents, or the fucking tactical agents or whatever other fucking fuckery. All i need is the fucking destination, and where to look Maybe to shoot, nothing fucking else, everything else fucks up everything.
Fuck this shit is frustrating to deal with.
Fuck this shit is frustrating to deal with.
There has to be some type of navigation system, otherwise the agent wouldn't be able to move to the destination at all. What you are referring to I think is a steering behavior implementation. Even that requires some type of navigation. You are the first to request this feature and I've added it to my list, but I do not have an ETA for it. The strategy would be the same as the A* integration with just implementing the base class. This is all that the NavMeshAgent version does, with the only difference being one is included in the Asset Store package and another being a separate package.
ill add my own navigation system. One which works for me No I'm not referring to steering behaviour. All i need from behaviour designer is a literal Vector3 destination, not steering, nothing. For the look-at tasks just tell me a vector3 look position, ill have my own system to make it look at there, some may use first person controller, some may use cars, i use a tank. The turret only needs to turn to LookAt direction, not the whole vehicle.
What i mean is... Behaviour designer telling some NavigationDemo_CS.cs the destination, the look at position and maybe a target to shoot at.
Then that NavigationDemo_CS.cs script has just a very simple navigation, where it just moves to the destination, looks at it etc. Super easy to remove script and replace with personal version.
Ill provide the vision etc, everything. i just need the behaviour tree to be the brains of vehicle and decide what to do with the information I'm providing.
Last edited: