Experimental AI - Revision history

Mattsc: /* Recruitment */

2022-06-22T23:14:55Z

‎Recruitment

Mattsc: /* Recruitment */ add [args] parameters

2022-06-20T15:36:35Z

‎Recruitment: add [args] parameters

Mattsc: /* Experimental AI Details */ update to current status in 1.15.3+dev

2020-07-19T18:15:22Z

‎Experimental AI Details: update to current status in 1.15.3+dev

Mattsc: /* Experimental AI Summary */ Update to current status in 1.15.3+dev

2020-07-19T18:11:07Z

‎Experimental AI Summary: Update to current status in 1.15.3+dev

Mattsc: /* Experimental AI Summary */ add a forgotten bullet

2019-12-11T20:56:26Z

‎Experimental AI Summary: add a forgotten bullet

Mattsc: Update after merge of ExpAI CAs into the default AI

2019-12-11T20:54:17Z

Update after merge of ExpAI CAs into the default AI

Mattsc: /* Custom Parameter Setup */ Fixed version (this is not going to be backported to 1.14)

2019-11-06T15:14:03Z

‎Custom Parameter Setup: Fixed version (this is not going to be backported to 1.14)

Mattsc: /* Experimental AI Setup */ Add custom parameters section

2019-10-17T14:04:11Z

‎Experimental AI Setup: Add custom parameters section

Mattsc: /* Experimental AI Setup */ Update ExpAI setup instructions

2018-12-25T22:46:56Z

‎Experimental AI Setup: Update ExpAI setup instructions

Mattsc: /* Experimental AI Setup */

2018-12-24T00:42:55Z

‎Experimental AI Setup

@@ Line 82: / Line 82: @@
 There is also a small amount of randomness added to allow two units that are almost equally good according to the evaluation to be chosen evenly, instead of only picking the first one every time.
-'''Configurable parameters:''' The [args] tag of the recruitment CA can take two parameters, '''high_level_fraction''' and '''randomness''' to modify its behavior.  They are described under the [[Micro_AIs#Rush_Recruitment_AI_.28ai_type.3Drecruit_rushers.29|Rush Recruitment AI]] entry on the Micro AI wiki page.
+'''Configuration parameters:''' The [args] tag of the recruitment CA can take configuration parameters '''high_level_fraction''', '''randomness''' and '''reset_cache_each_turn''' to modify its behavior.  They are described under the [[Micro_AIs#Rush_Recruitment_AI_.28ai_type.3Drecruit_rushers.29|Rush Recruitment AI]] entry on the Micro AI wiki page.
 ====Villages====

← Older revision		Revision as of 15:36, 20 June 2022
Line 81:		Line 81:

	There is also a small amount of randomness added to allow two units that are almost equally good according to the evaluation to be chosen evenly, instead of only picking the first one every time.		There is also a small amount of randomness added to allow two units that are almost equally good according to the evaluation to be chosen evenly, instead of only picking the first one every time.
		+
		+	'''Configurable parameters:''' The [args] tag of the recruitment CA can take two parameters, '''high_level_fraction''' and '''randomness''' to modify its behavior. They are described under the [[Micro_AIs#Rush_Recruitment_AI_.28ai_type.3Drecruit_rushers.29\|Rush Recruitment AI]] entry on the Micro AI wiki page.

	====Villages====		====Villages====

@@ Line 64: / Line 64: @@
 ====Recruitment====
-{{DevFeature1.15|3}} At the moment, the recruiting CA remains the main difference between the default and the Experimental AI. We are, however, currently working on adapting the Experimental AI's recruiting to work in more general settings (mostly this means with more than one leader), after which there will be a choice which recruiting to use in the default AI.
+{{DevFeature1.15|3}} At the moment, the recruiting CA remains the main difference between the default and the Experimental AI.
 The Experimental AI has a completely different recruiting algorithm that was designed to emulate the choices of a human player, especially in the first few turns.  As such, it tries to pick units that counter both what is on the battlefield and that the opponent could recruit.  As it gains more units relative to the enemy, it picks harder hitting units, even at the cost of fragility, to break through enemy lines.
@@ Line 86: / Line 86: @@
 The Experimental AI prefers to capture villages over fighting. This weakens it somewhat in raw fighting but it tends to get enough extra gold that this is actually to its advantage vs. the default AI, although its current implementation is probably too focused on villages.
-{{DevFeature1.15|3}} Besides recruiting, this is the only CA that is different between the default and Experimental AIs. It is too simple to work in a general setting (and therefore not suited for the default AI) and there is a separate village grabbing CA in the default AI, so it does not make much of a difference. This CA needs to be either improved or removed.
+{{DevFeature1.15|3}} This CA has '''not''' been merged into the default AI. It is too simple to work in a general setting (and therefore not suited for the default AI) and there is a separate village grabbing CA in the default AI, so it does not make much of a difference. This CA needs to be either improved or removed.
 ====Healing and Retreating====
@@ Line 99: / Line 99: @@
 The leader will move to different keeps, providing a forward force. Usually this works very well, especially at the beginning of the game, but it keeps moving later, which sometimes might get it killed in some scenarios. In that case, it might be worth stopping this behaviour after several turn by removing this CA.
 ====Enemy Poisoning====

@@ Line 1: / Line 1: @@
 == Experimental AI Summary ==
-{{DevFeature1.15|3}} Most of the Experimental AI candidate actions (CAs) '''have been merged into the [[RCA_AI|default (RCA) AI]]''', see descriptions below for the individual CAs. The Experimental AI is therefore currently only available in debug mode in the MP computer player selection menu, as it is almost the same as the default AI. This may change in the future if new features are added.  In scenario WML code, the methods described below to choose the Experimental AI work as before (but, of course, there will not be much of a difference to the default AI).
+{{DevFeature1.15|3}} Several of the Experimental AI candidate actions (CAs) '''have been merged into the [[RCA_AI|default (RCA) AI]]''', see descriptions below for the individual CAs.
 An ''Experimental AI'' is available for use in both MP maps and SP scenarios.  At the moment, this AI contains the following [[Wesnoth_AI_Framework#The_Candidate_Actions_of_the_main_loop_Stage|candidate actions]]:

@@ Line 6: / Line 6: @@
 * New and improved recruiting
 * More aggressive village grabbing
-* Healer placement behind injured units
+* Healer placement behind injured units and more assertive retreating of injured units to healing locations.
 * Leader castle switching: useful on certain MP maps
 * Poison spreading
 * All the [[RCA_AI#The_Candidate_Actions_.28CAs.29_of_the_main_loop_Stage_in_the_RCA_AI|default RCA AI candidate actions]] except for recruiting

@@ Line 1: / Line 1: @@
 == Experimental AI Summary ==
 An ''Experimental AI'' is available for use in both MP maps and SP scenarios.  At the moment, this AI contains the following [[Wesnoth_AI_Framework#The_Candidate_Actions_of_the_main_loop_Stage|candidate actions]]:
@@ Line 60: / Line 62: @@
 ====Recruitment====
 The Experimental AI has a completely different recruiting algorithm that was designed to emulate the choices of a human player, especially in the first few turns.  As such, it tries to pick units that counter both what is on the battlefield and that the opponent could recruit.  As it gains more units relative to the enemy, it picks harder hitting units, even at the cost of fragility, to break through enemy lines.
@@ Line 81: / Line 85: @@
 The Experimental AI prefers to capture villages over fighting. This weakens it somewhat in raw fighting but it tends to get enough extra gold that this is actually to its advantage vs. the default AI, although its current implementation is probably too focused on villages.
-====Healing====
+{{DevFeature1.15|3}} Besides recruiting, this is the only CA that is different between the default and Experimental AIs. It is too simple to work in a general setting (and therefore not suited for the default AI) and there is a separate village grabbing CA in the default AI, so it does not make much of a difference. This CA needs to be either improved or removed.
 If a healer is not used to attack, the Experimental AI will try to heal units, instead of positioning to attack later.
-Units retreat to villages/healers when injured, but currently there is little attempt to identify when it would be better not to in order to finish off multiple enemies in an area.
+In a separate CA, units retreat to villages/healers when injured.
 ====Castle Switching====
-The leader will move to different keeps, providing a forward force. Sometimes this works very well, especially at the beginning of the game, but it keeps moving later, which sometimes gets it killed. It might be worth stopping this behaviour after turn 3-4 or trying to identify the best keep rather than picking a close one that is not the current one as a way to improve it.
+The leader will move to different keeps, providing a forward force. Usually this works very well, especially at the beginning of the game, but it keeps moving later, which sometimes might get it killed in some scenarios. In that case, it might be worth stopping this behaviour after several turn by removing this CA.
 ====Enemy Poisoning====
@@ Line 95: / Line 105: @@
 There is also an explicit poison attack routine that tries to poison unpoisoned units instead of repeatedly poisoning the same one.
-====A Final Note====
+{{DevFeature1.15|3}} This CA has been merged into the default AI.
-* Many of the extreme biases in the Experimental AI come from the fact that it still uses a lot of the default AI code and thus cannot ask it what the default action would be and compare possible moves. It has to decide whether the alternative is better without knowing the default choice.
+{{DevFeature1.15|3}} This CA has been merged into the default AI.
 '''See also:''' [[Wesnoth AI]]
 [[Category:AI]]

@@ Line 38: / Line 38: @@
 === Custom Parameter Setup ===
-{{DevFeature1.14|10}} '''(and from 1.15.2 in current master)'''
+{{DevFeature1.15|2}}
 A small number of custom parameters can be passed to the Experimental AI. These cannot be set up using the <code>ai_algorithm</code> syntax, but require the use of a macro in the <code>[side][ai]</code> tag:
   [side]

@@ Line 35: / Line 35: @@
           #switch_ai=ai/ais/ai_experimental.cfg # Wesnoth 1.15 and later
       [/modify_side]
 == Experimental AI Details ==

@@ Line 11: / Line 11: @@
 == Experimental AI Setup ==
-In multiplayer maps, this AI is available from the game setup menu as 'Experimental AI'.  In single-player scenarios, it can be included by using the following code in a [side] tag<s>
+In multiplayer maps, this AI is available from the game setup menu as 'Experimental AI'.  In single-player scenarios, it can be included by using the following code in a [side] tag
   [ai]
       {EXPERIMENTAL_AI}
   [/ai]
 {{DevFeature1.15|0}} The EXPERIMENTAL_AI Macro is deprecated now. The Experimental AI should be included as follows instead (this works in Wesnoth 1.14 already):
   [ai]
       ai_algorithm=experimental_ai
   [/ai]
-'''Important Notes:'''
+Note that the EXPERIMENTAL_AI macro must be inside <code>[side][ai]</code>, it does not work with <code>[modify_side]</code> or <code>[modify_ai]</code>. If the <code>ai_algorithm</code> version is used, the Experimental AI can also be added as
-* This [ai] tag must go into a [side] tag.  It does not work with [modify_side] or [modify_ai].
+     [modify_side]
-* Currently, there is a restriction in the Wesnoth source code that makes it impossible to add the [[ AiWML#List_of_AI_Aspects|default aspects]] together with the Experimental AI inside the same [side] tag, even if they are in different [ai] tags.  We are working on lifting that restriction.  In the meantime, the Experimental AI must be defined in a [side][ai] tag as shown above.  Aspects can then be added by using [[Modifying_AI_Components#Using_.5Bmodify_side.5D_to_Change_Aspects_Mid_Scenario|[modify_side][ai]]] or [[Modifying_AI_Components#Modifying_Standard_Aspects|[modify_ai]]] in an event.
+         side=2
 == Experimental AI Details ==