rsunley
Posts: 33
Joined: 11/3/2009
From: Perth, Western Australia
Status: offline
|
I've been musing about the AI lately and wanted to offer up some thoughts. I've read the old posts about decision makers and so on but strictly in terms of strategy I have tried to come up with an abstract framework.
I see 2 main things that the AIO must be able to do. Select, prioritise and achieve goals and manage threats (although this could itself be a continuous goal). At a lower level, there will be actions that need to take place to accomplish those things but I do not want to get to that just yet.
For each threat
Prioritise - None, Minimal, Low, Moderate, High
- For each level of threat there will be a set of conditions that define it
- For each level of threat there will be a set of actions to mitigate the threat (maintaining the level) or reducing the level
For each goal
Prioritise - 1(high) - 5 (low). Higher priority objectives are tackled before lower priority ones.
- For each priority level there will be a set of conditions that define it
There will be a set of pre-requisites that must be present before the goal can be assessed
There will be a set of conditions that must be fulfilled in order to set the goal as achieved
There will be a set of conditions that define the goal as being unachievable (which would influence other goals)
In terms of politics, there is a posture against each major/minor power
Neutral
Conquered
Contain
Degrade
Conquer
For each power/level there are a set of goals. The level can change depending on progress towards those goals.
Certain postures may vary at the start of each game, for example the AIO may start with a plan to Conquer the CW (in which case a certain build strategy is required from the outset) or merely Degrade it.
I believe this approach could build a logical but predictable opponent. However real players are not always logical and can learn from their mistakes. Perhaps by introducing some variability in the setting of goal priority you could a) keep the human player on their toes and b) measure the priority settings against the achievement of goals so that with more play, optimal priorities are more likely to be chosen. This would require some kind of report to be sent back to Matrix for analysis. So as an example if the AIO is trying to degrade the CW, capturing Malta would be a goal (towards controlling the Mediterranean) , but over 100 games, how did that affect the CW ability to achieve its goals? And we wouldn't want the AIO to do this every game (unless it was undefended), there would be occasions where the focus would be to go hard for Alexandria instead.
I continue to chew on this matter and may post more later on
Ralph
|
Printable Version
Strategic objective framework - 
New Messages
No New Messages
Hot Topic w/ New Messages
Hot Topic w/o New Messages
Locked w/ New Messages
Locked w/o New Messages
Post New Thread