Simulation and Games - Looking for Information

You could ask me (nicely) about the whopper collection of algorithms that I have constructed to simulate airline passenger behavior in my game, but I''m not sure I can tell you about it yet!

Dave Mark - President and Lead Designer
Intrinsic Algorithm - "Reducing the world to mathematical equations!"

Is the main technique in simulation games merely to take the actual systems, model them into rules and processes, and then create simulated "people" that go through the rules and processes? If that''s the case, then would I also be safe in saying that the two main focuses simulation in games would be to create realistic and detailed processes, as well as creating simulated "people" that react realistically and make decisions within those processes?

I have one more related question: does the greatest difficulty in creating a simulation for use in a game result from creating the customers or people that go through the processes, or from creating the algorithms to model the processes, or both?

Each game design has its own difficulties so that is hard to answer in blanket form. However, in the games that you mentioned, i.e. SimCity and Rollercoaster Tycoon, much of the difficulty would be in recreating reality in a variety of forms. In RCT, for example, there is the recreation of physics for the coasters, a reasonable method of converting stimulus into reaction (e.g. "intensity", "excitement", "nausea") and then contstructing the people''s decision making process around those variables.

Dave Mark - President and Lead Designer
Intrinsic Algorithm - "Reducing the world to mathematical equations!"

There are two main approaches to modelling social and economic systems: process based and agent based. The process based approach tries to model the key parameters of the system by a set of equations. So, for example, a model of the stock market would attempt to model stock prices, or market indicators, utilising information such as the number of shares issues, number available, number of buyers, number of sellers, etc, etc,... A process model for the interactions of a group of people would be harder to define, but might model things like clique sizes, who-knows-who, information passes around, etc. I know that''s a little nebulous, but I''m trying to highlight the fact that the process model is an approximation of the system dynamics made by an outside observer. The process model implemented may have absolutely no bearing to the true underlying dynamics of the system, except for the fact that it should give similar results given an input state.

An agent based approach though seeks to define agents such that when these agents interact (according to the defined actions each agent can take) the end effect is a social or economic system (or something similar). There was some excellent work done on such systems at the Sante Fe Institute during the late 80s and early 90s. You might want to look there for references.

There is also a game in development at the moment (although I don''t recall the title) that utilises the agent based approach for interactions within the game. It''s based heavily on psycho-social and behavioural models of humans. I came across it earlier this year in a Gamasutra feature article, so that would be the place to start looking.

I hope this helps, at least a little, in your endeavour.

Cheers,

Timkin

Agent based relies heavily on emergent behavior and, therefore, can be a bit touchy.

Dave Mark - President and Lead Designer
Intrinsic Algorithm - "Reducing the world to mathematical equations!"

I''m just trying to make sure I understand the difference between agent-based and process based systems.

Take Rollercoaster Tycoon, for example. Does it use an agent based or process based approach? From looking at the game, I would have to guess agent-based, since it seems to be based on the idea of little people wandering around the theme park and interacting with various rides. Am I misconstruing the concept of the agent-based approach?

Innocuous Fox: What approach would you say that you used/use in your game?

A lot of times you are going to use both. In RCT, there is a process based model using formulas to calculate the stats for the coasters. The people are little agents wandering around interacting with their environment seemingly of their own accord based on their own preferences.

For my game, I am also using both. I have to model a lot of complex information such as generation curves to create realistic passenger traffic. However, once created, each passenger, airline, aircraft, gate and even runway is reacting and interacting with the other agents based on a series of behavior models that I have put into place for each agent.

Dave Mark - President and Lead Designer
Intrinsic Algorithm - "Reducing the world to mathematical equations!"

One final sorta related question: Simulation is not considered a part of the field of artificial intelligence, correct? So was I wrong to post my question in this forum? Just for future reference in case I have any more questions concerning the subject.

[edited by - ares32585 on July 31, 2003 6:53:17 PM]

That depends on what it is you are talking about... In a lot of ways, I believe the two are related. However, much of it depends on what you are modeling. If you are modeling the mathematics behind business decisions or the stock market, in a way that IS AI in the sense that you are trying to determine a way to algorithmicly and formulaticaly recreate something which is normally driven by human intelligence. If you are trying to recreate the physics of a bouncing ball, that is not necessarily AI in the classic sense, although AI programmers are often called on to resolve the physics of such things as bouncing balls, line of sight checks, etc.

Dave Mark - President and Lead Designer
Intrinsic Algorithm - "Reducing the world to mathematical equations!"

quote:

---

Original post by InnocuousFox
If you are modeling the mathematics behind business decisions or the stock market, in a way that IS AI in the sense that you are trying to determine a way to algorithmicly and formulaticaly recreate something which is normally driven by human intelligence.

---

I''d disagree with Dave on this one... partly...

A simulation of AI might involve creating a set of simulated agents and allowing them to interact according to their design (agent function). If they were intelligent, then you have simulated intelligent agents. On the other hand, simulating the effects of agent interaction is not AI. For example, in Dave''s Airline Game, passenger levels on flights are simulated using a series of mathematical models to determine numbers, rather than permitting individual agents to choose which flight they want, get stuck in traffic, get bumped, take a later flight, miss their connection, take a flight on another airline, etc, etc,... Generally speaking, a process model approach is not AI. However, that doesn''t mean an agent based approach is necssarily AI. It''s simply that one of the problems tackled in AI is to make simulated agents that behave like humans and/or act intelligently. Of course, since intelligence is still not clearly defined, most AI researchers just go with creating agents that act rationally, which is well defined.

As to the issue of emergent behaviour... well, yes, much of the nonlinearity of agent interactions is emergent and difficult to predict. However, a well designed system will behave appropriately on average .

Finally, Dave, the adverb of formula is formulaically , derived from the adjective formulaic .

Cheers,

Timkin

quote:

---

Original post by Timkin
For example, in Dave''s Airline Game, passenger levels on flights are simulated using a series of mathematical models to determine numbers, rather than permitting individual agents to choose which flight they want, get stuck in traffic, get bumped, take a later flight, miss their connection, take a flight on another airline, etc, etc,...

---

Actually, that is not entirely correct regarding my game. I use formulas to generate the passengers with origins, destinations and 6 different preference ratings for such things as cost, comfort, length of trip, etc. After that, they are a viable object that is processed individually based on those preferences. They book flights on their own, get on and off planes on their own, wait out delays and rebook if they have been stranded someplace. Depending on the preferences in place for that agent, they may handle things differently than the "person next to them."

quote:

---

As to the issue of emergent behaviour... well, yes, much of the nonlinearity of agent interactions is emergent and difficult to predict. However, a well designed system will behave appropriately on average .

---

One of the things you have to address when designing this sort of model is that you have to cover as many situations that the agent may encounter as possible... and then turn them loose. Emeregent behavior can be exciting as it unfolds... or your entire system may collapse on itself as your agents all end up with certain indicators pinned to the wall rather than flowing freely.

quote:

---

Finally, Dave, the adverb of formula is formulaically , derived from the adjective formulaic .

---

Great... I have a bloody Ausie teaching me how to speak.

Dave Mark - President and Lead Designer
Intrinsic Algorithm - "Reducing the world to mathematical equations!"

So the part of a simulation that would (possibly) be considered artificial intelligence would be the process of giving agents in the simulation the ability to make decisions and react to situations that are brought about by the modeled processes of the system being simulated?

quote:

---

Original post by ares32585
One final sorta related question: Simulation is not considered a part of the field of artificial intelligence, correct? So was I wrong to post my question in this forum? Just for future reference in case I have any more questions concerning the subject.

---

Ares, i personally don''t consider simulation AI. There''s no formal definition of AI and the field is a lot more nebulous than most, but here''s the definition i use - AI is what people who call themselves AI people are doing. Most people who call themselves AI people do not work on simulations. There is a different degree (Modeling & Simulation) in which almost everyone does simulations. So based on that definition, i''m gonna say that simulation isn''t AI. At a minimum, it''s not one of the top 10 research areas in AI

That having been said, if you didn''t ask the question on an AI newsgroup, where would you ask it? i figure this is where you should ask

i would love to hear how SimCity did their AI (or simulation). i''m sure someone must have written about it by now. But i don''t know any articles personally

SimCity, like Populous, Dungeon Keeper, Hotel Tycoon, etc. are considered god games. If you search for "god games" (the two terms together - you can use quotes in Google), you might have fewer bad hits than simulation of business or other search terms

Another option is to search for the names of the people who wrote the games you like and see if you find an interview. Who were those people? Wasn''t it Will Wright and Peter Molyneaux?

A few schools offer modeling PhDs. The big one is ucf.edu - University of Central Florida. One big thing they do is CGF, which is modeling people for RTS games. Not what you''re looking for. But maybe they also do the type of sims that can help you. i personally haven''t seen anything on their site, though, that i think would help

As for how the actual sims you enjoy work, i don''t know but i''m guessing it''s pretty easy. After all, these things ran on computers made of stone with monitors consisting of a parrot pecking out an image on a stone tablet. They had less processing power than a lethargic hamster and they ran the games fine so it can''t have been too bad (in the most simple case). But how did they do it?

Well, having now put 30 seconds thought into it, i''ll tell you how i would do it. First, recognize that the games i''m thinking of are grid based, so i''m going to treat a grid (city block or whatever) as the basic unit. Each unit consumes and creates. You define the what on both of those

Units can be modified. A grid cell can have electric poles, roads, curves and houses. One type of object that might or might not belong to the grid are the agents, things like cars and people. Let''s assume they are independent for now and are seeking the best offer

So what does a unit eat and produce? i''d say a typical unit consumes nothing. A unit + house consumes electricity and produces a signal (pheremone, fulfillment of desire, etc.) saying it''s a good place to live. If true, the unit will send out the signal "live here, i rock" if and only if it''s fed its inputs (here, electricity). So now the little mobile agents want to go there. Only the pathfinding will be a bit tricky because there are no roads. So the unit is inaccessible. Add roads and then what? The unit is somewhere and wants to go to the best place it can. So it looks at its options and if you''re the best place it drives to you. If everyone goes there it might get a lot of traffic which will produce a negative "live here, it rocks" signal which will make the unit look less desirable than other units. And if you have a power outage (because the nuke plant couldn''t produce enough energy to feed all the units) the unit stops being desirable

And so on and so on. You treat the world in terms of units, units have outputs only when they have inputs, units can attract or repel the wandering people (who have to come from somewhere; that one i don''t know) and after that i personally would just start tweaking the resource consumption and strength of outputs until the game felt right. But that''s me

Hope that helps a little

-baylor

quote:

---

Original post by InnocuousFox
Actually, that is not entirely correct regarding my game.

---

My apologies for misrepresenting your game. I clearly misunderstood what you wrote. Sorry.

quote:

---

Great... I have a bloody Ausie teaching me how to speak.

---

Well, since you yanks decided to invent your own dictionary, someone needs to remind you that you''re out of step with the rest of the planet!

Timkin

quote:

---

Original post by Timkin
Well, since you yanks decided to invent your own dictionary, someone needs to remind you that you''re out of step with the rest of the planet!

---

Although non-US people probably won''t care, it''s worth pointing out for the important people that there is the USA (the yanks) and then there are the truly important people (the Texans). As everyone (of significance) knows, Texas controls the United States and participates in it purely out of curiosity and boredom

i went on a trip to Europe a while back. Friends back home used to think that Texans were the most arrogant people on Earth until i came back with photos of the Texas embassies in London and Paris. That''s right, we have our own embassy. Do any other states have embassies? Heck no. But we do because we are special

So if it seems like people in the US are inventing their own spellings (like Brit spellings minus all the superferously colourful ''U''s) and using correct pronunciations, it''s only because we''re better. Remember, our king [u]is [/u] the Leader of the Free World

Some day you will all abandon the metric system and come to the dark side

-b, the definition of the ugly american

*sigh*

Dave Mark - President and Lead Designer
Intrinsic Algorithm - "Reducing the world to mathematical equations!"