TSI 0.24 beta released! (compatible with Spring 0.74b3)
Moderators: hoijui, Moderators
- Tow_IMPono
- Posts: 12
- Joined: 21 Aug 2006, 23:58
It took much more time than we expected (over two months), but finally TSI 0.24 is finished.
It corrects all known bugs of 0.21 version and brings several new features. Also a few new mods (like Kernel Panic) are supported.
TSI plays much better, but there is still much to be done.
Changelog:
This version generates more logs than 0.21 (yes, it's possible). If you don't want them, just delete or rename AI\TSI\log subfolder.
If you encounter any crashes or other problems we will be very grateful for reports.
>DOWNLOAD HERE<
(TSI package at UF will be updated by Tow Dragon ASAP)
Next releases:
TSI 0.3 will have full support for water maps. The very basic support is more or less ready, but due to lack of time it won't be finished (and released) very soon.
We are also going to rewrite unit management system, it'll be probably done in 0.4 version.
------
Umrug -> Thanks for patience. It is always rewarded :)
It corrects all known bugs of 0.21 version and brings several new features. Also a few new mods (like Kernel Panic) are supported.
TSI plays much better, but there is still much to be done.
Changelog:
AI\TSI folder must exist (else TSI will crash).0.21 -> 0.24 - 17.IV.2007
Fixed bug precluding TSI from building anti-air.
TSI won't crash any more on metal maps
Improved mex placing on metal maps
Improved scouting management
Light calibration
Support for new mods like Kernel Panic
Fixes bug causing freezes on maps with water
HeightMap - new powerful feature, with possibilities farly exceed anything form other AIs (however it will be fully used in TSI 0.4 or later)
New Mex placing algorithm
Fixed bug with geothermal power plants
A lot of minor fixes
This version generates more logs than 0.21 (yes, it's possible). If you don't want them, just delete or rename AI\TSI\log subfolder.
If you encounter any crashes or other problems we will be very grateful for reports.
>DOWNLOAD HERE<
(TSI package at UF will be updated by Tow Dragon ASAP)
Next releases:
TSI 0.3 will have full support for water maps. The very basic support is more or less ready, but due to lack of time it won't be finished (and released) very soon.
We are also going to rewrite unit management system, it'll be probably done in 0.4 version.
------
Umrug -> Thanks for patience. It is always rewarded :)
- tow_dragon
- Posts: 45
- Joined: 05 May 2006, 13:53
I'm surprised I'm the first to post test results here as AFAIK there are several people testing AIs nowadays... anyway...
- Only tested with BA47 on "Archers Valley" and "Ashap Plateau"
- Commander: never DGUN, rarely even shoots his laser (??? why is that, maybe because of endless e-stall? or stays out of range?)
- Commander: once hit by any skirmisher, engages in usual builder behavior, that is, runs away. Thats obviously bad as it'd better kill the puny kbot and keep the construction going.
- BIG problem: only builds wind generators thus having a terrible e-stall once in a few minutes, thats a totally first priority to fix
- goes good in fast expanding and capturing mexes, but first tiny kbot built by an enemy goes and kills all the expansion - need either to send attacking units to guard builders or built static defenses near each and every mex (thats always good for AI, I think, an LLT or two near every mex)
- in total, does not build enough static defences
- Attacking:
- Units are sent one-by-one at what seems to be random areas inside enemy base. It this scouting? With artillery vehicles? Even this could work when spamming, but units do not stop to attack enemy structures, but go endlessly to die inside enemy base.
- not enough attacking units at any stage of the game
:) Hope to see next version from you soon
- Only tested with BA47 on "Archers Valley" and "Ashap Plateau"
- Commander: never DGUN, rarely even shoots his laser (??? why is that, maybe because of endless e-stall? or stays out of range?)
- Commander: once hit by any skirmisher, engages in usual builder behavior, that is, runs away. Thats obviously bad as it'd better kill the puny kbot and keep the construction going.
- BIG problem: only builds wind generators thus having a terrible e-stall once in a few minutes, thats a totally first priority to fix
- goes good in fast expanding and capturing mexes, but first tiny kbot built by an enemy goes and kills all the expansion - need either to send attacking units to guard builders or built static defenses near each and every mex (thats always good for AI, I think, an LLT or two near every mex)
- in total, does not build enough static defences
- Attacking:
- Units are sent one-by-one at what seems to be random areas inside enemy base. It this scouting? With artillery vehicles? Even this could work when spamming, but units do not stop to attack enemy structures, but go endlessly to die inside enemy base.
- not enough attacking units at any stage of the game
:) Hope to see next version from you soon
- tow_dragon
- Posts: 45
- Joined: 05 May 2006, 13:53
Thanks for testing our AI ;].
Commander uses his D-GUN very rarely because of (probably) energy stall (which occures quite often due to incomplete economy handling) and high probability of destroying something important (especially in late game). Mayby we will work on it.
Wind generators are build always when it seems to be profitable (average wind power more than 8, AFAIR). It shouldn't be a big problem.
Better attacking units management is planned for version 0.4 (currently in development).
Commander uses his D-GUN very rarely because of (probably) energy stall (which occures quite often due to incomplete economy handling) and high probability of destroying something important (especially in late game). Mayby we will work on it.
Wind generators are build always when it seems to be profitable (average wind power more than 8, AFAIR). It shouldn't be a big problem.
Better attacking units management is planned for version 0.4 (currently in development).
Shouldnt the decision to make wind power be made from calculation of other available power production methods and their cost vs. power produced.AF wrote:Consider raising that value from 8 to 14
This way it would work for a bigger set of mods, and not relying in the ta style where solar gives +20
Example: if solar would give +5 and the cost would be same as currently in BA, then it would be insane to build solars even if the wind would be only 1-2. The AI could by a few simple calculations determine this and build wind.
- tow_dragon
- Posts: 45
- Joined: 05 May 2006, 13:53
I looked into the code and found algorithm that calculates the most efficient source of power (it's different form this one I mentioned, but I forgot that it had been changed).
Function that calculates the most efficient source of power returns the power plant that has the greatest value of following function:
(Let's ud be the pointer to unit's UnitDef)
(ud->energyMake - ud->energyUpkeep + min(averageWindPower*0.8, ud->windGenerator ))/(ud->energyCost+ud->metalCost*averageEnergyCostOfUnitsInThisModDividedByAverageMetalCostOfUnitsInThisMod)
Function that calculates the most efficient source of power returns the power plant that has the greatest value of following function:
(Let's ud be the pointer to unit's UnitDef)
(ud->energyMake - ud->energyUpkeep + min(averageWindPower*0.8, ud->windGenerator ))/(ud->energyCost+ud->metalCost*averageEnergyCostOfUnitsInThisModDividedByAverageMetalCostOfUnitsInThisMod)
There is no doubt that wind generator, given wind is above 0, is preferable to solars using this formula.
It looks like also min wind should be taken into consideration.
Ideally, it should build some of these and some of the others. Also, it's definitely worth for builders to build advanced solar (in BA).
It looks like also min wind should be taken into consideration.
Ideally, it should build some of these and some of the others. Also, it's definitely worth for builders to build advanced solar (in BA).