Some suggestions for recon and others.
Posted: Sun Nov 17, 2019 3:08 am
So far I mostly played large scale (15km x 15km map or at least 10km x 10km map + 37000pt ~ 50000pt) games only. I really had a fun> I really try to make AI more strong and try to set up AI behavior as streamlined as possible to prevent piece meal attack. I still see some problems, but unit-per-tile update really helped a lot.
But what I think it would be great to see is enhanced recon infantry AI. I'm not sure how recon infantry AI is coded at current version, but it seems they are mostly static and behaves like very passive & immobile observation point. However, as a person who has an experience of service in actual recon unit, this is not the only behavior recons did during operation, at least from my own experience during exercises I participated. There are several missions recons had to finish, but maybe I can introduce some suggestions here:
1. AI recon infantry infiltration algorithm.
How about infiltration algorithm for AI recon infantry. for recon infantry:
-A. Pick one VP
--a) In meeting engagement or defense game, pick the farthest one from friendly side.
--b) In offense, pick closest one from friendly side.
-B. From the circular region of within n-meter radius from VP, find the wood or structure tile for maximum concealment. Let's say n=500m for beginning but it should be an variable value with map size. While doing so:
-C. Count the total number of wood or structure tiles within the n-meter radius from VP zone. If it is less then certain % (=minimum surface area of concealment tile ratio) of n-meter radius zone, then increase the n number every 10m, check again. Repeat this until algorithm finds the satisfied n-meter radius. Repeat this check until n reaches to m (= maximum possible criteria for checking, variable depends on map size)
-D. If satisfied condition is found and n-meter radius zone from VP is satisfied, then let AI recon infantry attempt to infiltrate to the corresponding wood or urban tiles. While doing so:
-E. To prevent "forest island tile" or "small town tile" which is surrounded by open field and guarantees the failure of infiltration, there should be additional condition algorithm to check the total surface area of forest / urban tile, which is continuous from the forest / urban tile of #C and #D (forest / wood tile within n-meter radius zone from VP)
-F. However if all conditions are not fit: then this recons cannot infiltrate for OP of corresponding VP. In this case, go back to #A and try next VP.
-G. For whatever reason if all conditions are not satisfied: (desert map so no forest / urban and no high-concealment region to exploit, or too small map) then don't use infiltration.
In this case, the algorithm should rely on subfunction of "major routes or roads" among VPs. Considering the current AI mostly behaves based on VP position, I assume that AB has some algorithm to find 1st, 2nd, and 3rd priority roads / routes from one VP to next VP. Do #B ~ #F, but in this case everything should be based on the center point location of such road connection, not VP.
-H. If all are fail, (no roads connection, like totally middle of desert) then:
--a) In meeting engagement or offense game, do not use infiltration for AI recon infantry, but use advance / scout to VP of enemy side.
--b) In defense game, use the current algorithm for AI infantry: static and passive, work as static OP point.
-I. Move recon vehicles to advance to the good concealment tile near successfully infiltrated / moved AI recon team's location (within 10 tiles).
-J. Use infiltrated recons as early warning and FO element to call artillery and mortar support. Depending on the game, it would be possible to make AI to mix FASCAM and/or illumination.
Such use of recon infantry as early warning artillery FO will counter player's "rush to VP in opening" tactics. In meeting engagement, player usually rush to important VP faster than AI and make meeting engagement easy. Those early warning recon infantry will make such tactics really hard. In defense game, well infiltrated AI recon team would be possible to crush player's force before they reached line of departure or attack position. In offense game, well infiltrated AI recon team will more actively spot player's defense forces from concealed position and call artillery.
2. Spatial analysis / route analysis algorithm (I'm sure there is algorithm for these in AB, but just in case I add)
I'm curious, if this AB has any spatial analysis / route analysis algorithm for given map. IMO AI should set up attack position and line of departure (if there are no such algorithm for AI). This should be one of the ways to prevent (or at least reduce) piecemeal attack. Those 2 locations (attack position and line of departure) should be out of LoS from enemy position.
-1) By setting attack position, let AI forces to reduce the gap among elements. Let the spearhead AI units and others to stop from attack position until all (or 80% of) following units reached within specific radius. Then make AI forces assault together.
-2) If AI forces attempt pincer move by 2 or 3 directions, setting of line of departure would be great. Halt all AI forces when they reach the line of departure, and make them all assault together. This way, 2~3 groups of AI forces can harmonize their assault at the same timing, not piecemeal attempt which easly countered by player.
I think spatial analysis algorithm would be helpful to find the best location of lone of departure and attack position, and assault / advance route. At least those should be placed in the tiles of "cover" path algorithm. Or, it would be great to make a new criteria as "least visible tile" algorithm. For any given tile, count the total number of visible tiles based on LoS check. AI would be able to analzye and locate the "least visible area" based on "least visible tile" algorithm data. Those "least visible area" should be chosen as attack position or line of departure. Also, "least visible area" algorithm would help AI to find the best assault route (not cover path algorithm to move) to attempt suprise attack on player position. This "least visible area" algorithm will prevent AI to mostly rely on road to assault or advance. To me, current AI rely on road network a lot. But sometimes this makes game too easy. I wish AI to push through the woods not using the local road in the forest, but by push through the non-road field tiles. I'm sure such AI behavior will make player panic time to time.
3. "Commanding HQ" unit on the field
It does not matter if such unit is player HQ or not. However, I think there should be at least one "commander HQ", which is at least one level higher than all HQs. This will make this game's OOB a little bit realistic then current. Give huge penalty to command delay, artillery/airstrike arriving time, or anything regarding communication, if "commander HQ" is dead. If "commander HQ" is player HQ then that will be game over.
4. HQ hunt / protect algorithm for AI recon infantry.
With "Commanding HQ", it would be possible for recon AI to locate and call artillery or locate and directly attack to destroy "Commanding HQ", or any other HQs if not possible. On the other hand, AI recon infantry or vehicles would form a defensive perimeter around "Commanding HQ" to prevent player's attempt to assassinate "Commanding HQ".
North Korean KPA operates significant number of SF forces. https://fas.org/nuke/guide/dprk/agency/ ... part03.htm. They call all of them as SF, but considering the logistics capability of NK, those "SF forces" will mix of light infantry, somewhere between light infantry ~ SF, and real SF.
One of the prime mission of KPA recon infantry team (usually such light infantry ~ SF team) is, hunt down and eliminate RoKA (Republic of Korean Army or South Korean Army) HQ elements in the AO. They are so good at it, as they have very good stamina, morale, training degree, and movement speed. Once they defeat ROKA or US HQ elements in the AO, the command structure of RoKA or US forces will be disrupted so that any ROKA / US operations will be stalled or significantly delayed. When I was active, our major goals was detect and prevent infiltration of such KPA recons, light infantry, and SFs. On the other hand, I was also participated in some sort of raiding team to track and hunt down OPFOR HQ elements in AO during exercise. Considering that the most of KPA doctrines were (and still are) mostly copy and paste of Soviet / PLA doctrines, I guess such "HQ hunt" should be one of the priority mission list for recon / light infantry / SF for cold war armies. (But this part is my guess)
For these operations, SF and light infantry elements should be introduced in all factions of AB in recon tab. I think "tier 1 SF" would hardly join "conventional operations" but any lower tier SF and light infantry would have a bit better chance to join "conventional operations", such as GRU Spetnaz, Green Berets and Rangers, Airborne or air-assault divisions, or Luftsturmregiment...
Such "HQ hunt / protect" mission will make AB more interesting + increase one more tactical / strategical layer to players and AI + increase difficulty for players and bring balance. "HQ hunt / protect" mission will be a bit different from the goal of "conventional approach". Rather, it will be a bit in a direction of COIN/guerrilla operation, so any faction with significantly inferior armament (such as KPA) will have better chance to win or turn the tide if such operation become successful.
I think developing an algorithm for locate / search and call fire support or direct assault wouldn't be difficult, as I'm thinking several idea. Protection of friendly HQ for AI would be also easy.
But what I think it would be great to see is enhanced recon infantry AI. I'm not sure how recon infantry AI is coded at current version, but it seems they are mostly static and behaves like very passive & immobile observation point. However, as a person who has an experience of service in actual recon unit, this is not the only behavior recons did during operation, at least from my own experience during exercises I participated. There are several missions recons had to finish, but maybe I can introduce some suggestions here:
1. AI recon infantry infiltration algorithm.
How about infiltration algorithm for AI recon infantry. for recon infantry:
-A. Pick one VP
--a) In meeting engagement or defense game, pick the farthest one from friendly side.
--b) In offense, pick closest one from friendly side.
-B. From the circular region of within n-meter radius from VP, find the wood or structure tile for maximum concealment. Let's say n=500m for beginning but it should be an variable value with map size. While doing so:
-C. Count the total number of wood or structure tiles within the n-meter radius from VP zone. If it is less then certain % (=minimum surface area of concealment tile ratio) of n-meter radius zone, then increase the n number every 10m, check again. Repeat this until algorithm finds the satisfied n-meter radius. Repeat this check until n reaches to m (= maximum possible criteria for checking, variable depends on map size)
-D. If satisfied condition is found and n-meter radius zone from VP is satisfied, then let AI recon infantry attempt to infiltrate to the corresponding wood or urban tiles. While doing so:
-E. To prevent "forest island tile" or "small town tile" which is surrounded by open field and guarantees the failure of infiltration, there should be additional condition algorithm to check the total surface area of forest / urban tile, which is continuous from the forest / urban tile of #C and #D (forest / wood tile within n-meter radius zone from VP)
-F. However if all conditions are not fit: then this recons cannot infiltrate for OP of corresponding VP. In this case, go back to #A and try next VP.
-G. For whatever reason if all conditions are not satisfied: (desert map so no forest / urban and no high-concealment region to exploit, or too small map) then don't use infiltration.
In this case, the algorithm should rely on subfunction of "major routes or roads" among VPs. Considering the current AI mostly behaves based on VP position, I assume that AB has some algorithm to find 1st, 2nd, and 3rd priority roads / routes from one VP to next VP. Do #B ~ #F, but in this case everything should be based on the center point location of such road connection, not VP.
-H. If all are fail, (no roads connection, like totally middle of desert) then:
--a) In meeting engagement or offense game, do not use infiltration for AI recon infantry, but use advance / scout to VP of enemy side.
--b) In defense game, use the current algorithm for AI infantry: static and passive, work as static OP point.
-I. Move recon vehicles to advance to the good concealment tile near successfully infiltrated / moved AI recon team's location (within 10 tiles).
-J. Use infiltrated recons as early warning and FO element to call artillery and mortar support. Depending on the game, it would be possible to make AI to mix FASCAM and/or illumination.
Such use of recon infantry as early warning artillery FO will counter player's "rush to VP in opening" tactics. In meeting engagement, player usually rush to important VP faster than AI and make meeting engagement easy. Those early warning recon infantry will make such tactics really hard. In defense game, well infiltrated AI recon team would be possible to crush player's force before they reached line of departure or attack position. In offense game, well infiltrated AI recon team will more actively spot player's defense forces from concealed position and call artillery.
2. Spatial analysis / route analysis algorithm (I'm sure there is algorithm for these in AB, but just in case I add)
I'm curious, if this AB has any spatial analysis / route analysis algorithm for given map. IMO AI should set up attack position and line of departure (if there are no such algorithm for AI). This should be one of the ways to prevent (or at least reduce) piecemeal attack. Those 2 locations (attack position and line of departure) should be out of LoS from enemy position.
-1) By setting attack position, let AI forces to reduce the gap among elements. Let the spearhead AI units and others to stop from attack position until all (or 80% of) following units reached within specific radius. Then make AI forces assault together.
-2) If AI forces attempt pincer move by 2 or 3 directions, setting of line of departure would be great. Halt all AI forces when they reach the line of departure, and make them all assault together. This way, 2~3 groups of AI forces can harmonize their assault at the same timing, not piecemeal attempt which easly countered by player.
I think spatial analysis algorithm would be helpful to find the best location of lone of departure and attack position, and assault / advance route. At least those should be placed in the tiles of "cover" path algorithm. Or, it would be great to make a new criteria as "least visible tile" algorithm. For any given tile, count the total number of visible tiles based on LoS check. AI would be able to analzye and locate the "least visible area" based on "least visible tile" algorithm data. Those "least visible area" should be chosen as attack position or line of departure. Also, "least visible area" algorithm would help AI to find the best assault route (not cover path algorithm to move) to attempt suprise attack on player position. This "least visible area" algorithm will prevent AI to mostly rely on road to assault or advance. To me, current AI rely on road network a lot. But sometimes this makes game too easy. I wish AI to push through the woods not using the local road in the forest, but by push through the non-road field tiles. I'm sure such AI behavior will make player panic time to time.
3. "Commanding HQ" unit on the field
It does not matter if such unit is player HQ or not. However, I think there should be at least one "commander HQ", which is at least one level higher than all HQs. This will make this game's OOB a little bit realistic then current. Give huge penalty to command delay, artillery/airstrike arriving time, or anything regarding communication, if "commander HQ" is dead. If "commander HQ" is player HQ then that will be game over.
4. HQ hunt / protect algorithm for AI recon infantry.
With "Commanding HQ", it would be possible for recon AI to locate and call artillery or locate and directly attack to destroy "Commanding HQ", or any other HQs if not possible. On the other hand, AI recon infantry or vehicles would form a defensive perimeter around "Commanding HQ" to prevent player's attempt to assassinate "Commanding HQ".
North Korean KPA operates significant number of SF forces. https://fas.org/nuke/guide/dprk/agency/ ... part03.htm. They call all of them as SF, but considering the logistics capability of NK, those "SF forces" will mix of light infantry, somewhere between light infantry ~ SF, and real SF.
One of the prime mission of KPA recon infantry team (usually such light infantry ~ SF team) is, hunt down and eliminate RoKA (Republic of Korean Army or South Korean Army) HQ elements in the AO. They are so good at it, as they have very good stamina, morale, training degree, and movement speed. Once they defeat ROKA or US HQ elements in the AO, the command structure of RoKA or US forces will be disrupted so that any ROKA / US operations will be stalled or significantly delayed. When I was active, our major goals was detect and prevent infiltration of such KPA recons, light infantry, and SFs. On the other hand, I was also participated in some sort of raiding team to track and hunt down OPFOR HQ elements in AO during exercise. Considering that the most of KPA doctrines were (and still are) mostly copy and paste of Soviet / PLA doctrines, I guess such "HQ hunt" should be one of the priority mission list for recon / light infantry / SF for cold war armies. (But this part is my guess)
For these operations, SF and light infantry elements should be introduced in all factions of AB in recon tab. I think "tier 1 SF" would hardly join "conventional operations" but any lower tier SF and light infantry would have a bit better chance to join "conventional operations", such as GRU Spetnaz, Green Berets and Rangers, Airborne or air-assault divisions, or Luftsturmregiment...
Such "HQ hunt / protect" mission will make AB more interesting + increase one more tactical / strategical layer to players and AI + increase difficulty for players and bring balance. "HQ hunt / protect" mission will be a bit different from the goal of "conventional approach". Rather, it will be a bit in a direction of COIN/guerrilla operation, so any faction with significantly inferior armament (such as KPA) will have better chance to win or turn the tide if such operation become successful.
I think developing an algorithm for locate / search and call fire support or direct assault wouldn't be difficult, as I'm thinking several idea. Protection of friendly HQ for AI would be also easy.