Menu
▾
▴
#50 Campaign AI behavior specialties
Since Richard began working on new AI for Single Player campaigns about 3 months ago, it seemed expedient to me to try to exchange my own outlooks here early on, in hope that more useful ideas might come into sight along the way.
To begin with, there definitely were certain weak points in the original AI's behavior which were impelled more by considerable hardware limitations which existed in that time than the desire of Dune 2 developers to go down the road of not utilizing more than 1 Harvester for the AI, no MCVs to rebuild lost Construction Yards at all, the total limitation of only 81 buildings which can exist at the same time on the level, and, almost everlasting "Unable to Create More" units message which popped up on the last ones.
While the reasons for them were quite obvious, I have to admit it's sometimes rather hard to try to fit all design decisions one might want to apply to levels arranged in that old fashion, with the goal of making them more consistent, challenging and compatible with both game versions.
Here's the list of things I personally find not challenging enough in the original AI's behavior:
1) The utilization of only 1 Harvester,
2) The inability to restore Construction Yards through deployment of a new MCV,
3) The absence of any logics for management of Starport invoices,
4) Very little utilization of building captures via Sabotage missions,
5) And, generally poor understanding of where the player base's weak points are.
The AI of Legacy's engine is designed rather well, however, it sometimes does things the original (non-Super Dune 2) levels simply don't grant enough credits for, for example, to prevail over 5 "AI Player (Hard)" opponents, one would normally need about 15.000 credits and at least 6 Harvesters, 5 Launchers and 6 Siege and normal Tanks available at Starport, along with some Ornis.
I'd say it simply adheres to the motto "The best defense is offense", and it does so no less efficient than a very experienced player. Is it good? Yes. But, is it as much entertaining for that game mode as it's beneficial for the winner? Hardly so, as inputting everything into offense spawns rather plain, one-sided type of 25 minutes blitz-play. Yes, these opponents can be defeated, but, will there be a lot of fun in having almost-similarly to do so? Always the same 25-30 minutes, otherwise, you're likely to run out of Spice, or, you'll simply get overridden by the harvesting defensive groups the enemy will have near your base.
Here are some of the things in which I deem the right route of improvement:
1) Pre-placed units, which have been stationed by map designers to protect bases of computer players from scenario's beginning should be replaced over time in case they've been destroyed. If there are no longer any buildings belonging to the house controlling them within "Area Guard" range from the unit's initial stationing point, the position is considered not needing a replacement until the next check in 5 minutes. If nothing has been rebuilt for 20 minutes, the position is considered abandoned and is not checked further.
2) These pre-placed units shouldn't be normally participating in defensive missions or attacks made upon the player, unless under special conditions, in which the group's refitting script will be allowed to take lacking units from those guarding the base, but, in this case they still have to be replaced over time with exactly the same type of unit which existed here in "Area Guard" mode prior to attacking/defending task reallocation.
The mission itself is the factor which defines whether pre-placed unit is considered protective or not. If it wasn't appointed to reside in "Area Guard" mode, it's not treated as one, thus not activating necessity of replacement.
3) The limit of utilized Refineries should be set to 3, while each of them must be supposed to host 2 Harvesters, resulting in 6 total per AI player, for which at least 3 Carryalls should be maintained in possession.
As the end result, the Campaign AI is supposed to represent a balanced combination of offensive, riposting and defensive measures, making it even better than in the original game, and yet, not exercising determinative blitz-type logic.