GSoC Akihara Proposal - Revision history

Esr: Hide historical references to Subversion from the wiki search function

2013-03-20T23:59:29Z

Hide historical references to Subversion from the wiki search function

Akihara: /* Current Code? */

2012-08-23T14:43:11Z

‎Current Code?

Akihara: /* Scoots and village */

2012-08-23T13:38:12Z

‎Scoots and village

Akihara: /* Scoots and village */

2012-07-29T19:19:02Z

‎Scoots and village

Akihara: /* Scoots and village */

2012-07-29T19:00:41Z

‎Scoots and village

Akihara: /* Scoots and village */

2012-07-29T18:48:32Z

‎Scoots and village

Akihara at 18:45, 29 July 2012

2012-07-29T18:45:32Z

Akihara at 14:54, 26 July 2012

2012-07-26T14:54:17Z

Akihara: /* Important area */

2012-07-15T12:39:12Z

‎Important area

Akihara: /* Important area */

2012-07-15T12:38:04Z

‎Important area

@@ Line 726: / Line 726: @@
 Moreover, I coded in c, c++, python, prolog and I think it's all.
-  * Subversion - I have my own svn.
+  * Sub&shy;&shy;version - I have my own s&shy;&shy;v&shy;&shy;n.
   * C++ - I developed in c++ some years ago but I will remember since the start of GSoC
   * STL, Boost, Sdl  - never used.

@@ Line 717: / Line 717: @@
 When we choose to recruit one unit, we must not forget to taken it into account for the next one. Otherwise, AI will choose the same unit indefinitely.
-==== Current Code? ====
+== Wanna test some functions? ==
 = Practical considerations =

← Older revision		Revision as of 13:38, 23 August 2012
Line 640:		Line 640:

	Of course, certain villages can be ommited bu this algorithm. One behind the castle for example.		Of course, certain villages can be ommited bu this algorithm. One behind the castle for example.
		+
		+
		+	'''Limitations'''
		+	We can ameliorate this algorithm into may ways:
		+	* First, the villages recognition. Infact, some villages can be inaccessible (for example, a village surrnouded by any unwalkable field). In this case, the village might be ignored. For this, we can use a ''pathfind'' to determine the movement cost.
		+	* As an extention of the first point, some field aren't walkable for certain unit, or some other can be faster. It might be interresting to know them.

	===== Important area =====		===== Important area =====

@@ Line 611: / Line 611: @@
 In code format, these information will be stored in a int according to a enum type:
-  enum searchDirection { UP = 0, DOWN = 1, LEFT = 2, RIGHT = 4 }
+  enum searchDirection { UP = 1, DOWN = 2, LEFT = 4, RIGHT = 8}
 The values are chose in order to have a unique value when adding one to an other.
 = UP + LEFT
 = DOWN + LEFT
 = UP + RIGHT
 = DOWN + RIGHT
 = DOWN + RIGHT + LEFT
 * Searching villages
-We will search villages according to the searching direction. In the case of our example, the castle is in position 15 - 2 with a searching direction equal to 7. We will searching all the village which are in the searching zone (all the village under the castle in this case) and sort them according to the distance between them and the castle.
+We will search villages according to the searching direction. In the case of our example, the castle is in position 15 - 2 with a searching direction equal to 14. We will searching all the village which are in the searching zone (all the village under the castle in this case) and sort them according to the distance between them and the castle.
 This distance will be a 'dummy' distance, we won't ake the field into account.
@@ Line 630: / Line 631: @@
   Location: 11 - 4
-  Searching Direction: 3
+  Searching Direction: 6
 This mean that we need to find villages which have a location as x <= 11 and y >= 4. In our example, we are lucky! One village is in the zone: location 5 - 9. It will be our next step :) We redo the same thing on this village until we found no more village in the zone. We have one path :) Then, we choose the nearest no-marked village and do it again.

@@ Line 617: / Line 617: @@
 = UP + RIGHT
 = DOWN + RIGHT
 * Create paths
@@ Line 624: / Line 632: @@
   Searching Direction: 3
-This mean that we need to find villages which have a location as x <= 11 and y >= 4. In our example, we are lucky! One village is in the zone: location 5 - 9. It will be our next step :) We redo the same thing on this village until we found no more village in the zone. We have one path :) Then, we choose the nearest no-marked villageand do it again.
+This mean that we need to find villages which have a location as x <= 11 and y >= 4. In our example, we are lucky! One village is in the zone: location 5 - 9. It will be our next step :) We redo the same thing on this village until we found no more village in the zone. We have one path :) Then, we choose the nearest no-marked village and do it again.
 In the case of the village in area 3, we can see that they are two village in the searching zone. We always will chose the nearest one!

← Older revision		Revision as of 18:48, 29 July 2012
Line 609:		Line 609:

	If a castle is in area 1 for example, the searching direction will be RIGHT and DOWN. In this case, the village position don't matter for the direction :) We always use the castle as our first research position.		If a castle is in area 1 for example, the searching direction will be RIGHT and DOWN. In this case, the village position don't matter for the direction :) We always use the castle as our first research position.
−
−	~~If you don't really understand the direction settings according to the castle and villages position, take a look at the picture above.~~
−
−	~~IMAGE!~~

	In code format, these information will be stored in a int according to a enum type:		In code format, these information will be stored in a int according to a enum type:

@@ Line 580: / Line 580: @@
 All the algorithm need to be done once in order to have the final list. When we decide to recruit a unit, the only thing to do is to cancel unit we cannot recruit anymore according to gold.
-===== Scoots and village =====
+==== Scoots and village ====
 Scoots are very special unit. They are usually used to take villages around . The problem here is "How many scoot should be recruit?".
@@ Line 591: / Line 591: @@
 At least, we will have two zones to explore: zone A and B. The two of them will be explored in the same way.
-We will seek to create a graph on which each vertex will be a village. To simplify the algorithm, and also make more sense, we won't consider ALL villages. In fact, nobody could take all villages before the opponent. It makes same to consider the 60% of them regarding to our castle.
+'''For the moment, we will forget the notion of BlueZone'''. It will come later :)
-The number of scoots necessary will be N.
+Let's take an example !
-'''How to determine the number of path
+http://ar.chroot-me.in/random/wesnoth/scout_example.png
-For the moment, we will forget the blue zone. So, we have one unique zone.
+This screen is a part of the ai_arena map. There is a lot of informations on this picture, but we will take them one by one.
-* '''Choose villages'''
+* The '''crisscross'''
-For choosing villages, we will build a dummy distance calculating between two points. With this, we will have this informations for example:
+* The '''searching direction'''
-We are on Castle A :) As a default value, we will take into account villages which respect: Da <= 60% of (Da+Db). So we will have a list of correct villages. A good thing to do is to have a sorted list with the nearest village as the first element.
+If we look at the picture, the castle is located into area 2 which is the most particular area (with 4,6 and 8). Indeed, they are 'middle' areas, so the direction ''searching direction'' depend of the village location. Let's take the example of the village on area 2, and the nearest one in area 3. They don't have the same ''search direction'' (green area) !!
-* '''Crisscross the field'''
+Well, the explanation is simple: the castle is in area 2, a ''middle'' area. So the ''searching direction'' depend of the village location. Let's imagine that area 2 is divided into two parts verticaly: villages which are on the right of the castle will have DOWN and RIGHT for searching direction, wheareas villages on the left will have DOWN and LEFT.
-The aim of crisscrossing the field is the determine specified areas. We will crisscross it for having nine areas. These areas wll determine which villages choose next :)
+If a castle is in area 1 for example, the searching direction will be RIGHT and DOWN. In this case, the village position don't matter for the direction :) We always use the castle as our first research position.
-At the end, we will have our N paths.
+If you don't really understand the direction settings according to the castle and villages position, take a look at the picture above.
-'''How to determine the blue zone'''
+In code format, these information will be stored in a int according to a enum type:
-The blue zone represent the straight line between two castle with a width we will fix. With the location of the two castle, it would be easy the determine the straight line formula.
+ enum searchDirection { UP = 0, DOWN = 1, LEFT = 2, RIGHT = 4 }
 ===== Important area =====

← Older revision		Revision as of 14:54, 26 July 2012
Line 596:		Line 596:

	The number of scoots necessary will be N.		The number of scoots necessary will be N.
		+
		+	'''How to determine the number of path
		+
		+	For the moment, we will forget the blue zone. So, we have one unique zone.
		+
		+	* '''Choose villages'''
		+
		+	For choosing villages, we will build a dummy distance calculating between two points. With this, we will have this informations for example:
		+	Da: Castle A - village distance = 12;
		+	Db: Castle B - village distance = 8;
		+
		+	We are on Castle A :) As a default value, we will take into account villages which respect: Da <= 60% of (Da+Db). So we will have a list of correct villages. A good thing to do is to have a sorted list with the nearest village as the first element.
		+
		+	* '''Crisscross the field'''
		+
		+	The aim of crisscrossing the field is the determine specified areas. We will crisscross it for having nine areas. These areas wll determine which villages choose next :)
		+
		+	At the end, we will have our N paths.

@@ Line 611: / Line 611: @@
 '''''But do we do next?'''''
-When we have determine these areas, we can determine a percentage of battle for a couple of terrain type. For example, ''att->flat vs def->swamp'' with a number equal to 2. It means that we can have 2 battle with these caracteristics on the map. Warning: the ''def->flat vs att->swamp'' is count as the same battle.
+When we have determine these areas, we can determine a percentage of battle for a couple of terrain type. For example, ''att->flat vs def->swamp'' with a number equal to 2. It means that we can have 2 battle with these caracteristics on the map. Warning: the ''def->flat vs att->swamp'' is count as the same context.
 Structure used for storing these informations: