CGF Terrain Specific AI

Like many tactical 1^st person shooters, Action Quake2 (notably AQ2 team play) heavily rewards tactical skills. Having a good aim and quick responses alone don't make you a successful player. Instead, it is mandatory to understand:

• what weapon to use where and when,
• when to rush, to hide, to reload, to provide suppression fire, or to bandage
• communication and roles within the team
• where to go, camp, snipe, move, or ambush

Of these four groups of skills, the 'where' understanding is one of the easier skills for humans to acquire (even our cats seem to be born with it). And there is plenty of literature available providing additional rules and guidelines (army field manuals, strategy guides for games, etc.).

For AI, however, the 'where' understanding is hard to implement efficiently (AI stands for artificial intelligence, or (in this case) programming of the bot 'brains'). For example:

• recent (top-selling) tactical games such as Rogue Spear display realism in their modeling of hostage, environment and weapons, but have troubles fitting maneuvers to the terrain. In the Rogue Spear demo, the default AI plan and behavior has the 'blue team' traverse a big museum hall in close formation through the center of the hall, thus becoming a potential victim for attacks from all angles while being unable to reach cover quickly. (See image above) ¹.

The tactical 'where' understanding is a very essential part of CGF: teams are expected to traverse all of the terrain autonomously and intelligently using solely plain waypoints (that are devoid of 'AI' hints such as 'regroup here', 'spread out', etc.).

Initial CGF betas included already some terrain understanding. However, CGF 0.80 will include revamped terrain understanding AI. Subsequent pages will illustrate the 'what and how' behind CGF terrain understanding.

The following terrain understanding is present in CGF (items will be completed in the coming weeks). Click on one of the items to continue:

1. terrain analysis (ranking positions for tactical value
   • analyis of static terrain, thus assessing all relevant terrain positions for tactical (dis)advantages
   • analyis of terrain usage, thus assessing where and how combat activities occur, and improving the 'static terrain' analysis results with this input
   
   
2. tactical path planning [implemented, to be written]
   • find paths that offer a stealthy approach
   • find paths that offer cover from enemy fire
   
   
3. stop planning [implemented, to be written]
   given a path, determine best path locations to
   • assemble/regroup
   • provide suppression fire from
   • check 'six'
   
   
4. threat movement prediction [implemented, to be written]
   upon losing contact with a threat,
   • predict where threats will be/move to
   
   
5. 'next position' attack planning [implemented, to be written]
   within a team, determine for a member
   • location to attack from
   • location to find cover (bandage, reload, ambush, disengage)

While on the topic of RB6 Rogue Spear: judging from the demo, it's a very interesting game, making up for many of the problems RB6 had (no crouching, no interaction with the environment (breakable glass)). Multi-player RS games are a bit like AQ2 games, but your movement is still limited (try jumping from the balcony 3 meters down - you cannot).