AI Improvements (Dev Diary)
Firefly Studios’ Designer Stephen Richards and series creator Simon Bradbury have revealed four major improvements to AI behaviour found in their upcoming ‘castle sim’ sequel Stronghold: Warlords. This new entry in the historical real-time strategy series will usher in major improvements and refinements in how computer controlled opponents interact with the player, each other and any warlords present on the map.
AI behaviours related to castle management, skirmish combat and city building have been completely revamped for Stronghold: Warlords, in addition to being able to work with the new warlords system. Much like the archetypes of the warlords themselves – such as the Ox, Tiger and Crane – each enemy AI will adhere to a distinct style of play over a set of predetermined behaviours. The decisions made by computer opponents will now be made according to map layout, tactical positioning of warlords, available resources, current army composition and their priorities at different stages of gameplay.
In practical terms this means less hard coded, predictable behaviours and more flexibility in how AI opponents can outmanoeuvre the player, but also how it deals with the evolving situation within its own castle. Independent of the player, AI will build, sell resources and train their army according to changing priorities over the course of a skirmish match. Just as a human player would, the AI will adjust its focus between city building and RTS combat while also battling it out with the player for control over each map’s selection of warlords.