MKelm
commented
10 years ago Swarm intelligence
I heard the term in a web project several years ago, and it was used in relation to user generated contents.
In relation to online features I talked about user generated content already. It can be attractive to use user generated content to create game modifications automatically and to collect the modification data on a server to redistribute them to the clients.
But the solutions are usable for an offline game too, the gamer can create his own game contents only then.
I have the idea to use it for:
- product ideas/design, if the basic concept of products contents functionality only and no content, the ideas and the design of them can be done by users, that allows more freedoms and the product catalog grows with the users/swarm, the products needs some meta data e.g. to make them available with the related technology levels for other gamers and for product design we can use a canvas painter tool like this jQuery plugin
- planet design, if we have a planet editor, we can use the user generated planets too, to use the users intelligence for that
- creating A.I. by swarm intelligence, we can record the decision chains of gamers automatically to create A.I. data for CPU companies, that makes it possible to make a game which have no A.I. data at the start, but it will be generated through the gaming phase, it needs some type of compatibility check, if some logic dependencies for A.I. data are broken in relation to a newer game version or not
An important element of swarm intelligence are rating/voting systems, to create meta data about the quality of an item, if it is useable (any longer). So each type of items need such a system, and we need an interface layer for it, it can be a layer with DOM objects, that would not look like the game interface directly, but it can make to design such an interface easier. It is more like the Steam layer in steam games.
Voting:
- planet/product items can use a catalog with voting/rating functions
- A.I. data by swarm needs some more data to train the global A.I. swarm intelligence (containing all sub data sets of single gamers) a catalog is not useable by gamers, it needs more an automatic re-distribution system of data, with a probability that the gamer gets some more untested A.I. data or less untested A.I. data, in every case the game needs another type of user vote in relation to opponents actions / strategies, but that is hard to realize, if the gamer does not spy the opponents the whole time, so how he gets to know which strategy / decision node the opponent chose and if it was too easy or too hard for the gamer?
A possible solution for swarm A.I. votes: Maybe a solution is to simulate market surveys in the game, that the gamer gets the role of a consultant for this feature, and a market analysis company sends him surveys every ingame year. The survey contains a random selection of (short) articles or notes with some information about different companies and their strategies in the last year, and the user can vote if he thinks it was a good strategy for the market or a bad one. E.g. a company which uses strategies which give it a high market position (monopoly) over a longer time is bad for the market.
I start this issue with a link to some tutorials to create Dungeon/Monster A.I. http://www.reddit.com/r/gamedev/search?q=tinykeep
That is not an AI the game needs directly but maybe it has some general mechanism which can be used.
Our monsters are up to 49 mass control companies on each planet (map). To make it easier I think they can not leave their map to trade and expand, and they do not know the planet map, they are more located on something like a decision node map. So they have to decide which decision way to choose, and each decision way can have branches and further ways until an intermediate goal is reached.
I read some sentences of the first monster article only yet, but there is something about different states of monsters, and we can add these states to, to make some decision ways more or less attractive. The decision node branches need something like a resistance (like in electronics) and the company owner has a percolation value which makes a node branch more or less attractive ...