Game (Simulation) Idea "Hide and Seek"

You should look up Nuns on the Run, I've never played it but it sounds very similar to what you're trying to do.

Hey, sorry for the slow response, school started today and I've been trying to really think about this.

Overall, I like most of your items. The only one I wouldn't super care for is the fake wall. I guess it just seems, I don't know, out of place. That said, this is just my opinion.

Could a player see the mud? If so, then why would he step in it if it could be avoided?

One item you might consider adding would be bubble-gum, tar, or something like that that would temporarily disable the player.

As to the overall structure of the game, when you described this at first, I pictured something slightly more old-timey. Like, instead of "win" or "lose," you play until you lose three lives and aim for the high score. Levels could get progressively harder, either because you programmed the AI to be harder, or because you used neural networks to allow it to out-think the player over time. I'm also not the hugest fan of getting points for hiding (although, if it is hide-and-seek, maybe that's fine), and would rather play a game where you get points for capturing minor flags/item, and then get a bunch for capturing the big one (of course, doing that brings you to the next, harder, level). If it was James Bond, it could be something like stealing secret plans and then making it to the exit or something. That said, this last bit is mostly my own personal taste, so feel free to completely disregard it.

///So as far as programming
If you could expand on your code structure, how is it learning, are you tracking its “intelligence” with a single variable and allowing it to focus more on the player based on that?

Are you tracking its intelligence based on many variables? For instance the bot knows that the last 6 games the player has started in the lower right corner of the map. Because of this the bots disperse in that direction (ultimately ignoring the opposing corner).

Variable = “Players are in the lower right corner”
Players start in position = x/y ++ VAR
Then of course you’d contrast this VAR against the VAR for the other corners.

Now, I’m not some code guru but I know my way around a compiler and theirs method to my madness. When creating war-games (my favorite) a variety of mechanics can be used to establish new kinds of game play. A perfect example of this is “active defense”.

If a YOU are about to take damage from my attack then YOU roll defense. This forces YOU to take on a certain sense of culpability for the death of your model. I may have killed it, but that’s hardly where most players’ heads are at after botching an “active defense” roll. Most players are either disappointed in the dice; or themselves. This kind of mechanic use serves both the purpose of lessoning animosity in a competitive game as well as allowing the defending player “YOU” a certain amount of responsibility for the models death.

If you could expand on how the AI works I think the best choice for a new and interesting mechanic would be allowing the player an active role in the AI. Just like “active defense” we could use the method of the AI itself as a mechanic. Methods of tricking or misleading the AI will give the player both a sense of accomplishment (push button, get bacon) as well as an ego boost (I’m smarter than it). Actively outsmarting the machine seems like the best goal for “Hide and Seek”.

In contrast to what others have said I think the worst approach would be simple mechanics such as obstructions. Much of this of course hinges on the kind of play style you’re looking for (fast or slow).

Love to hear what direction you decide to go with this.
E

Well lets take special exception to our potential reading audience. Ive become quite fond of my fellow BGDF'rs and would love to include them in this. Also, by virtue of thier inclusion I can illustrate my next point for possible direction in your design. One tidy little mess :)

So then, I do understand your apporach to this design. You, as i percieve it, are gathering together your thoughts on how youd like the game to be played. Your creating a vision of what it looks like as its played and how you want to create interactions between player and program. May I suggest a different approach?

For the most part your spot on. A clear vision should certainly be at the forefront of any design. In this case however your method for design will directly impact your end product. AI, in this instance, will evolve with the players. Because of the kind of function your attempting to get out of your bot I would have your change your perception of the project.

Your work should begin with the design of the AI. Put simply, this "world" revolves around a central component, the AI. By first defineing the AI and what parameters create its body we gain the ruleset that defines our "world".

As an example, if the AI can hear then the player could also make sound. This allows for a variety of mechanics to be implemented based on our "world"; which is ultimately an extension of attributes of the AI. In the example above we have defined sound as an attribute the AI can handle. If the AI can handle this attribute then we can also effect the attribute. As an example, a blind bot "type" could wander aimlessly listening. The player could then use a "boom" grenade to deafen this bot but would have to be careful not to expose himself to other bots who could of course see the explosion.

By predefineing the rules of the world we are better able to generate new mechanics and simplify our design method. Once parameters are defined for the AI the mechanics of the game become a simple matter of deduction and innovation. This design method is of course my own personal preference, I use this method almoast to the exclusion of all others in my own work.

By defineing attributes that the human element can associate with will give the player a certain sense of reality and offer strong immersion in a graphically lacking simulation. Try to think of players well enough immersed to profanely exclaim thier frustrations at the little blue dot which represents thier player. Players grumbling at thier machine because they "ran" just a tiny bit too much and, as a result, were more easily tracked by a "sniffer" bot. Now the player may have to work more on thier timeing, rather than useing a burst of speed, or take an alterate path. By offering attributes in your design the human element can associate with we can gain further immersion and thus encourage interaction.

If your interested in methods of implementing this kind of AI I can suggest some rudimentary methods of action tracking etc. The most critical item you should work on first is the AI's ability to deal with bump barriers. Make sure the AI moves cleanly and deals with walls intelligently. From this basic mechainc you will have actually crteated a huge portion of the infrastructure the AI will be relying on for its other attributes.

On a side note "swarm intelligence" is incredibly fun to work with and would add an entirely new dimension to the bots ability to track a player (as well as opening a massive can of worms as far as potential game mechanics). The scale of this project being defined would certainly help me make better choices when offering my two cents about implementation of mechanics.

E

Firstly, this sounds like a fun project to implement. I hope it goes well.

I agree with E that setting the rules first is the best way to approach the design.
As such, the list of rules you have listed is an excellent start. Your earlier posts presented many options which would vastly complicate the AI algorithm (using items, etc), which I doubt you would have time to implement properly in one semester.

By listing the rules, it highlights that you have at least one rule missing: namely a rule that explains why the bots should not just leave 1 bot sitting on the flag permanently. This is the best way to ensure that the bots win. As such, in order to be a believable AI (and to give the player an ego boost for "outsmarting it"), your goal for the AI should be to have it try to win using the best strategy that it can use within the ruleset. Otherwise when the player wins he thinks, "I only won because the AI was too stupid to leave a guard at the flag."
For example, add a rule that no bot may go within radius X of the flag (easily explainable via various themes, but the rule is the important part). Alternatively, add a rule that the AI does not know where the flag is (again, easily explainable via various themes).

There are several more assumptions / rules which you have probably got in your head which have not made it onto paper. Such as:
- What is the relationship between the player's movement speed and the bots? (same, faster, slower) The answer to this impacts on the player's strategy and the AI strategy.
- Do the bots have to touch the player to win, or do they just have to see him? Do they have to get within radius X of the player? Again, this affects the strategies available to both the player and the AI.

Now taking the player's perspective on what makes a "fun" game to play: waiting around is boring. As such, games where you win by waiting for the time limit to expire are generally boring. Similarly, earning points for doing nothing is boring. In this regard, your latest ruleset where the player wins by taking action and loses when the timer expires sounds like a much better game to play than the earlier suggestions.

Hope that helps,
kos

It sounds like the number of seekers will increase as the game progresses. If a seeker reported player position in an area, then this proposed map would explain how the seekers could form a dragnet around the player. This dragnet would be proceeded by additional seekers going into the dragnet and searching you out.

The first decision for a seeker would be whether or not to directly pursue the player rather than organize the trap. For that determination you might want to research Pursuit-Evader problems. There is a good paper by Starr&Ho that details the math involved.

Actually calculating the Dragnet Map would use the result from some of the previous maps. First you would need to define a player space based on the Prediction map. Then you would need a way of detailing all the ways out from the space through adjoining tiles. You want a solution where you select a minimum number of tiles with bad(not easy to hide) bottlenecks on the Hiding Spot Map. Theoretically this map could be created at export, with each solution grid associated with particular points on the Prediction/Spotted maps.

----------------------------------------------------------------------------------------------------

I agree with the poster who suggested the seekers do not know what the player's target(the flag) is. This fits in with security guards or base guards hunting down an intruder and solves the "guard the flag" issue.