New Games.lt Lithuanian interview

[Placebo 29]

Games.lt have kindly given permission for any fansite wishing to post the English version of their new interview with us to do so.

Proviso is that the interview must be posted in full and include credits to the site and writer from Games.lt along with clickable banner (included) which links to Games.lt

GamesLT_Interview.zip

Format of the interview can be changed as long as the content is not, also feel free to post in other languages

Anyone not respecting the request from me to post the interview with a clickable Games.lt banner runs the risk of not being included in future releases, which of course is something we'd all prefer to avoid

[Espectro (DayZ) 0]

English version available at:

http://www.armedassault.eu/readarticle.php?article_id=12

Thanks Games.lt and Placebo

[Guest Ti0n3r]

Quote[/b] ]Another area of AI we're working on at the moment is a system of leapfrogging/bounding overwatch which is a specialised military technique for combining tactical movement with suppressive fire with the aim of moving progressively forward whilst reducing the potential threat of enemy fire.

[martinovic 0]

Woot, lots of great new info! And just look at those new screens they look purdy! http://www.armedassault.eu/photogallery.php?photo_id=211

[berghoff 11]

Nice.

[colt 0]

It's very heartening to read about what BIS are trying to achieve with the AI (using trees, vehicles and the environment as cover and the leapfrogging). I hope they can pull it off.

What's a "FSM based modular AI system" ?

[Espectro (DayZ) 0]

It's very heartening to read about what BIS are trying to achieve with the AI (using trees, vehicles and the environment as cover and the leapfrogging). I hope they can pull it off.

What's a "FSM based modular AI system" ?

I asked placebo about that, he wouldn't tell me

[berghoff 11]

Maybe he doesn't know what it is aswell? =)

[Espectro (DayZ) 0]

This is an abstract of a paper i found:

Quote[/b] ]This paper describes a set of software tools created for constructing

and verifying a modular FSM for use as logic control.

Previously we developed a method for representing the logic of a

machining system by several modules. This method allows each

module to be designed and verified independently of the other

modules. Once the modules are verified to be correct, the system

can be verified to be deadlock free and absent of some undesired

behaviors, without having to construct a single large FSM representing

the entire system. These tools are written in Java to

facilitate integration with other software being developed for the

mechanical design of reconfigurable machining tools at the Engineering

Research Center for Reconfigurable Machining Systems

at the University of Michigan.

I guess it is a way to design AI groups to interact with other AI groups, and in that way create a dynamical enviroment for the entire AI force on an island indluenced by various factors of different sizes.

In short; a small firefight in the north of an island may affect AI forces on the south of the island. These other forces will then ask for back in their area affecting more forces creating a domino effect on all forces that are present in the island.

Did that make sense?

[Balschoiw 0]

Wild guess:

Quote[/b] ]Finite State Machines (FSM), also known as Finite State Automation (FSA), at their simplest, are models of the behaviors of a system or a complex object, with a limited number of defined conditions or modes, where mode transitions change with circumstance.

Finite state machines consist of 4 main elements:

* states which define behavior and may produce actions

* state transitions which are movement from one state to another

* rules or conditions which must be met to allow a state transition

* input events which are either externally or internally generated, which may possibly trigger rules and lead to state transitions

A finite state machine must have an initial state which provides a starting point, and a current state which remembers the product of the last state transition. Received input events act as triggers, which cause an evaluation of some kind of the rules that govern the transitions from the current state to other states. The best way to visualize a FSM is to think of it as a flow chart or a directed graph of states, though as will be shown; there are more accurate abstract modeling techniques that can be used.

FSM is typically used as a type of control system where knowledge is represented in the states, and actions are constrained by rules.

"...One of the most fascinating things about FSMs is that the very same design techniques can be used for designing Visual Basic programs, logic circuits or firmware for a microcontroller. Many computers and microprocessor chips have, at their hearts, a FSM."[1]

Finite state machines are an adopted artificial intelligence technique which originated in the field of mathematics, initially used for language representation. It is closely related to other fundamental knowledge representation techniques which are worth mentioning, such as semantic networks [5] and an extension of semantic networks called state space [5].

Semantic networks were proposed to represent meaning and relationships of English words. A graph is constructed where nodes represent concepts and edges the relationships. State space is an extension on the idea of semantic networks, where a node denotes a valid state and the edges transitions between states. State space, unlike FSM, requires both an initial state and a goal state, and is typically used in problem solving domains where a sequence of actions is required for solving the overall problem (sequence from initial to goal states). Like FSM, state space has rules which constrain state transitions, and are triggered by input events.

Like any rule based systems, if all the antecedent(s) of a rule are true, then the rule is triggered. It is possible for multiple rules to be triggered, and in the area of reasoning systems, this is called a conflict set. There can only be one transition from the current state, so a consistent conflict resolution strategy is required to select only one of the triggered rules to fire and thus performing a state transition.

This brings us to two main types of FSM. The original simple FSM is what’s known as deterministic, meaning that given an input and the current state, the state transition can be predicted. An extension on the concept at the opposite end is a non-deterministic finite state machine. This is where given the current state; the state transition is not predictable. It may be the case that multiple inputs are received at various times, means the transition from the current state to another state cannot be known until the inputs are received (event driven).

An implementation of a deterministic finite state machine may see the firing of the first rule that is triggered. This may be ideal for many problem domains, but for computer games, easily predictable behavior is usually not a wanted feature as it tends to remove the "fun-factor" in the game.

"...a player feels like they are playing against a realistic simulation of intelligence, and not against a reproduction of a sequence of actions." [2]

The "sequence" which is one of the key benefits of FSM, should not be blindingly obvious in computer games. There are a number of extensions to FSM and workarounds for "mixing up" the sequence to make it harder to predict actions. One of these non-deterministic approaches involves the application of another proven artificial intelligence technique; Fuzzy Logic, called Fuzzy State Machines (FuSM).

Just like finite state machines there is a lot of flexibility when implementing a fuzzy state machine. A fuzzy value can be applied to various state transitions. When a conflict set is encountered the higher the fuzzy value for a transition, the higher the likelihood of the state transition. This allows the specification of a fuzzy priority to state transitions.

An implementation of FuSM may involve the assignment of fuzzy values to various inputs to represent the degree an input is defined. The fuzzy system would use these weighted input values in the evaluation of rules, triggering only state transitions whose assessed value is above a specified threshold.

Another approach for converting a deterministic FSM into a non-deterministic FSM would be to simply use a random number generator to select a triggered rule. It may not be necessary to implement a deterministic finite state machine to have a perceived level of unpredictability. This can be achieved by a system or object that has a large number of defined states and a complex mesh of transitions, giving the appearance of being unpredictable.

It is important to understand the difference between a state and an action. When designing a computer program, larger functionality are decomposed into a number of smaller actions or activities. This is done so that each can be defined in a function, making the overall solution modular, and easier to maintain. FSM is similar in that it’s a decomposition of the behaviors of a system or object, and even a state can be decomposed into sub-states. The difference is a state may involve one or more actions.

Example 1: a moveUnit() action may be used by both the evadeEnemy state and the attackEnemy state.

Example 2: the evadeEnemy state may consist of many actions, some evaluations, some movement directives, and some actions which can change the entities own state. If the entity was cornered for example, there may be a state transition from evadeEnemy to attackEnemy, where the act of being cornered is the trigger.

The best way I like to think of the terms is an action is an activity that accomplishes something like an evaluation or a movement, and a state is a collection of actions that are used when in a particular mode. A state is the circumstance of a thing, its condition, and the actions are the attributes of that state. It provides the ability to limit the scope of actions or the amount of knowledge to only that required for the current state.

...

Full article:

Finite State Machines

Hope that helps

[D.murphy man 0]

Im especially happy about the new infomation on the AI. Glad to know the AI we've seen in recent videos wont be as stupid as in the final release (hopfully)

Now maybe we can have some decent urban combat too

[Espectro (DayZ) 0]

Im really looking forward to the "Frogleeping" technique, that I remember from my own time in service in here Denmark. It is very effective, and looks quite impresive if seen done right.

imagine a whole bunch of AI's using that technique

[NeMeSiS 11]

This is an abstract of a paper i found:

Quote[/b] ]This paper describes a set of software tools created for constructing

and verifying a modular FSM for use as logic control.

Previously we developed a method for representing the logic of a

machining system by several modules. This method allows each

module to be designed and verified independently of the other

modules. Once the modules are verified to be correct, the system

can be verified to be deadlock free and absent of some undesired

behaviors, without having to construct a single large FSM representing

the entire system. These tools are written in Java to

facilitate integration with other software being developed for the

mechanical design of reconfigurable machining tools at the Engineering

Research Center for Reconfigurable Machining Systems

at the University of Michigan.

I guess it is a way to design AI groups to interact with other AI groups, and in that way create a dynamical enviroment for the entire AI force on an island indluenced by various factors of different sizes.

In short; a small firefight in the north of an island may affect AI forces on the south of the island. These other forces will then ask for back in their area affecting more forces creating a domino effect on all forces that are present in the island.

Did that make sense?

Sounds like a build-in Guard waypoint

EDIT:

Quote[/b] ]It's not much use to be wearing a woodland camouflage uniform and to then be fighting in amongst beige coloured buildings,

A dynamic 'camouflage' setting, instead of the one in OFP?

And its great to hear that the are working on the AI, although we could have guessed it, its nice to be sure of it

[Kos'aaK 0]

Thanks for the translation, good interview and good news

[lwlooz 0]

Very nice interview,I especially like the AI part.

[figjam 0]

Great interview

Really nice to see the AI is getting the upgrade I was hoping for

[Talyn 33]

Very interesting read, particularly about the AI.

[anders^on 0]

Really interesting read, thanks for sharing.

We did a German translation at Armed-Assault.de:

German Version of the interview

[VladAlex 0]

Of course i agree ! It's a very fine interview. But all that is just words. We have absolutly no idea of what is going to be effective in the REAL game. We just dont know nothing... And the E3 demo proove that a lot of promess was very very far... So, I'm very sceptical...

[CorlanMcD 0]

Looks like ArmA is coming along just fine.

[alx 20]

That interview made me hope they implement proper vision limitations to the AI in dark/foggy situations. For instance, in Forgotten Battles (a WWII flight sim) the damn AI can see through clouds and just fine at night. With the dynamic weather being a feature, I'd hae to be slotted all the time by uber-AI.

[Zendjir 0]

Finally some info on the AI, sounds superb! I'm glad the guys at BIS share the communities vision on the importance of good AI.

[dreday 1]

Of course i agree ! It's a very fine interview. But all that is just words. We have absolutly no idea of what is going to be effective in the REAL game. We just dont know nothing... And the E3 demo proove that a lot of promess was very very far... So, I'm very sceptical...  

I was as critical of the E3 demo AI as anyone else on this board, perhaps even more... At the time there were some very vocal members of the board that had said that the AI was just fine the way it was; or that it could not be improved any further.

I am very impressed and grateful that BIS has chosen to listen to us and not to them. I don't know how many of the promised features we will see in the final product. However, I now have no doubt that BIS understands the importance of improved AI for overall game play; and that they will do their best to deliver on their promises.

IMHO, now is not the right time to be skeptical; now is the time to be supportive.

Peace,

DreDay

[bravo 6 0]

ho well.. another interview...

... ok..

..nice.

[CorlanMcD 0]

ho well.. another interview...

... ok..

..nice.

Get this man the MP demo quick! Before he passes out!.