User:Jleldridge - Revision history

Jleldridge: /* Demo */

2013-05-22T22:31:39Z

‎Demo

Jleldridge: /* Demo */

2013-05-22T22:30:31Z

‎Demo

Jleldridge: /* Demo */

2013-05-22T22:28:54Z

‎Demo

Jleldridge: /* Demo */

2013-05-22T22:27:03Z

‎Demo

Jleldridge: /* Demo */

2013-05-22T22:25:36Z

‎Demo

Jleldridge at 21:51, 22 May 2013

2013-05-22T21:51:00Z

Jleldridge: /* Detailed Description */

2013-05-21T18:40:04Z

‎Detailed Description

Jleldridge: /* Detailed Description */

2013-05-21T18:10:12Z

‎Detailed Description

Jleldridge: /* 4) Project */

2013-05-17T19:36:34Z

‎4) Project

Jleldridge: /* 4) Project */

2013-05-17T19:34:56Z

‎4) Project

@@ Line 42: / Line 42: @@
 Included here is a picture of the formation created by the total defense Micro AI around the villiage it is assigned to defend:
-<a href="https://github.com/jleldridge/wesnoth-old/blob/totaldefensedemo/total_defense_of_village.png">total_defense_of_village.png</a>
+https://github.com/jleldridge/wesnoth-old/blob/totaldefensedemo/
 Here is the commit for the demo. Every commit I made during the demo has been rebased into a single commit so that the related file diffs can be more easily viewed: https://github.com/jleldridge/wesnoth-old/commit/bb478407e29003a930a4cd395018adc7f36aa7c2

@@ Line 41: / Line 41: @@
 I also ran this scenario with just two healer support Micro AIs pitted against each other, with the northern team assigned a "protect_area" goal centered around the same village with a radius of 5 for comparison with the total defense Micro AI. In both scenarios the northern (defending) team was forced to stop recruiting after turn 4, so that the southern team would eventually overwhelm the northern team with the idea being to see how long each AI could keep the opponent from capturing the defended village. The performance of my total defense Micro AI in comparison to a simple protect_area goal was somewhat underwhelming. The team in general would last about the same amount of turns as the protect_area AI, though it is worth noting that the total defense Micro AI would never lose control of the village, while the protect_area AI would. Generally the leader would die due to not being a part of the defensive structure around the village in the total defense AI.
-Included here is a picture of the formation created by the total defense Micro AI around the villiage it is assigned to defend: ...
+Included here is a picture of the formation created by the total defense Micro AI around the villiage it is assigned to defend:
-https://github.com/jleldridge/wesnoth-old/blob/totaldefensedemo/total_defense_of_village.png
+<a href="https://github.com/jleldridge/wesnoth-old/blob/totaldefensedemo/total_defense_of_village.png">total_defense_of_village.png</a>
 Here is the commit for the demo. Every commit I made during the demo has been rebased into a single commit so that the related file diffs can be more easily viewed: https://github.com/jleldridge/wesnoth-old/commit/bb478407e29003a930a4cd395018adc7f36aa7c2

@@ Line 41: / Line 41: @@
 I also ran this scenario with just two healer support Micro AIs pitted against each other, with the northern team assigned a "protect_area" goal centered around the same village with a radius of 5 for comparison with the total defense Micro AI. In both scenarios the northern (defending) team was forced to stop recruiting after turn 4, so that the southern team would eventually overwhelm the northern team with the idea being to see how long each AI could keep the opponent from capturing the defended village. The performance of my total defense Micro AI in comparison to a simple protect_area goal was somewhat underwhelming. The team in general would last about the same amount of turns as the protect_area AI, though it is worth noting that the total defense Micro AI would never lose control of the village, while the protect_area AI would. Generally the leader would die due to not being a part of the defensive structure around the village in the total defense AI.
-Included here is a picture of the formation created by the total defense Micro AI around the villiage it is assigned to defend:
+Included here is a picture of the formation created by the total defense Micro AI around the villiage it is assigned to defend: ...
 https://github.com/jleldridge/wesnoth-old/blob/totaldefensedemo/total_defense_of_village.png

@@ Line 41: / Line 41: @@
 I also ran this scenario with just two healer support Micro AIs pitted against each other, with the northern team assigned a "protect_area" goal centered around the same village with a radius of 5 for comparison with the total defense Micro AI. In both scenarios the northern (defending) team was forced to stop recruiting after turn 4, so that the southern team would eventually overwhelm the northern team with the idea being to see how long each AI could keep the opponent from capturing the defended village. The performance of my total defense Micro AI in comparison to a simple protect_area goal was somewhat underwhelming. The team in general would last about the same amount of turns as the protect_area AI, though it is worth noting that the total defense Micro AI would never lose control of the village, while the protect_area AI would. Generally the leader would die due to not being a part of the defensive structure around the village in the total defense AI.
-Included here is a picture of the formation created by the total defense Micro AI around the villiage it is assigned to defend: https://github.com/jleldridge/wesnoth-old/blob/totaldefensedemo/total_defense_of_village.png
+Included here is a picture of the formation created by the total defense Micro AI around the villiage it is assigned to defend:
 Here is the commit for the demo. Every commit I made during the demo has been rebased into a single commit so that the related file diffs can be more easily viewed: https://github.com/jleldridge/wesnoth-old/commit/bb478407e29003a930a4cd395018adc7f36aa7c2

@@ Line 34: / Line 34: @@
 =Demo=
-https://github.com/jleldridge/wesnoth-old/commits/totaldefensedemo
 =IRC=

@@ Line 32: / Line 32: @@
 <h4>Rotate strong, undamaged units to replace damaged units.</h4>
 This would also be used in parallel with any of the other ideas. Units that make up the defensive formation should be rotated out when they are likely to die in the next turn or two. This would only occur if it is possible to replace that unit with a unit that is likely to live in that spot for longer than the current unit. This may be best implemented as a stand-alone candidate action using the cpp engine since it would likely be used by multiple defense strategies.
 =IRC=

@@ Line 29: / Line 29: @@
 <h4>Have the AI make an intelligent fallback when necessary.</h4>
 This would be developed in parallel to each of the other ideas, because it could be applied to each of them. The idea here is to sort out whether defending a specific area is actually possible or worthwhile based on the kind of force an enemy is attacking with. Heuristics could be used to measure if it is worthwhile to stay and sacrifice units to delay the enemy, or to retreat to save units. If a Micro AI were used, parameters could be passed in to modify how "hopeless" a situation would need to be before the AI abandoned the position. This mechanic would also specify where the AI should retreat to, and possibly include a heuristic to pick another currently defended or defensible position to fall back to.
 =IRC=

@@ Line 207: / Line 207: @@
 .5) Include as much technical detail about your implementation as you can.
-I plan to use the LUA engine and custom Micro AIs with high candidate action scores under certain conditions to exactly position units once they manage to get into the general are that is to be defended. Micro AIs allow for the grouping of several behaviors into one file, as well as the selection of units that will be affected by the Micro AI, so Micro AIs seem like the most flexible way to create total defense behavior.
+I plan to use the LUA engine and custom Micro AIs with high candidate action scores under certain conditions to exactly position units once they manage to get into the general area that is to be defended. Micro AIs allow for the grouping of several behaviors into one file, as well as the selection of units that will be affected by the Micro AI, so Micro AIs seem like the most flexible way to create total defense behavior.
 To define the area that is to be defended, "protect_unit" and "protect_area" [goal] tags could be used to modify where the movement candidate actions send units. The "protect_unit" and "protect_area" goal tags use the number of enemy units within a radius of the thing to be protected as a heuristic; if this proves insufficient to get units into desired defensible positions, new	goals can be created that use different heuristics to work in combination with the current goals. If goals are not flexible enough to get the desired movement behavior, a Micro AI can be used that explicitly ignores the movement candidate actions and takes care of moving units into defensible positions itself.

@@ Line 191: / Line 191: @@
 <b>July 15:</b>
 <ul>
-<li>Have more refined versions of the strategies, having determined which ideas work best in each</li>
+<li>Have more refined versions of the strategies, having determined which ideas work best in each engine.</li>
 </ul>