Welcome to the LnLP Forums and Resource Area

We have updated our forums to the latest version. If you had an account you should be able to log in and use it as before. If not please create an account and we look forward to having you as a member.

Command Ops 2 AI comparison possible?

Iconoclast

Member
Joined
Oct 26, 2014
Messages
130
Points
18
Location
germany
Hi all,

Long time no see!

I just stumbled across this post by Ezra Sidran who is developing his game called "General Staff". I was wondering if the CO AI works in similar ways, or if that would be something that we could expect in the future?

Cheers,

A
 
Joined
Oct 20, 2014
Messages
1,182
Points
63
Age
76
Location
Livonia, MI (Detroit-area suburb)
Hi all,

Long time no see!

I just stumbled across this post by Ezra Sidran who is developing his game called "General Staff". I was wondering if the CO AI works in similar ways, or if that would be something that we could expect in the future?

Cheers,

A
Having worked on projects that attempted to substitute artificial intelligence for "man in the loop" planning for the US Army, I can assure you that all AI efforts focus on the same objectives, and each approaches to those objectives on different paths.

As a SME for logistics integration into US Army land combat tactical training software (a project still under development at Sandia Labs), I can also assure you that CO2, despite the constraints of personal PC computing engines, has at least matched the quality and the accuracy of Army training software, and in some cases has exceeded it in establishing a viable AI-assisted combat model.
 

Iconoclast

Member
Joined
Oct 26, 2014
Messages
130
Points
18
Location
germany
Having worked on projects that attempted to substitute artificial intelligence for "man in the loop" planning for the US Army, I can assure you that all AI efforts focus on the same objectives, and each approaches to those objectives on different paths.

As a SME for logistics integration into US Army land combat tactical training software (a project still under development at Sandia Labs), I can also assure you that CO2, despite the constraints of personal PC computing engines, has at least matched the quality and the accuracy of Army training software, and in some cases has exceeded it in establishing a viable AI-assisted combat model.

Interesting!

I wasn't trying to pit GS and CO or any other programme against another. I just found it interesting to see "AI" under the hood, and for a title like CO, with probably the most infamous wargaming AI, I also thought it might be interesting to hear a bit from you guys how the AI in CO makes decisions.

Cheers,

A

Edit: See below
 
Last edited:

Dave 'Arjuna' O'Connor

Panther Games Designer
Joined
Jul 31, 2014
Messages
3,415
Points
113
Location
Canberra, Australia
Website
www.panthergames.com
Re "Infamous wargaming AI". That's a bit harsh. Maybe you meant "famous" and not "infamous".

Re AI decisions. The AI is what I refer to as "generic" in that it's not tied to a specific scenario or situation. Each unit has agency. It is aware of the situation around it and aware of the orders it has. When it receives orders it calls on its plan doctrine to develop courses of action and ultimately a plan of action. It allocates any resources under command to the various tasks in its plan and sends orders. When it commences a task it consults its task doctrine to determine what event it will execute next. Eg If it's a Move task then it will go through a series of assessments to ensure it can still move - ie it has enough fuel, it's not exhausted, everyone in the formation is where they should be etc. If it can, it schedules a new Move event for the next minute. It then executes its Move event.

After that it goes through a function called the ScheduleNextEvent(). This does a lot of the reaction and reassessment test. If none of these are activated then it will schedule the next event and so it will move along the road to its Move task's objective location. But if it encounters enemy it may well react by halting, deploying, firing, calling in fire support etc. It may reassess and change its route or abandon the task and bunker down. There are lots of different options considered every minute of the game. Once it arrives at its objective, it will determine if the task if fulfilled and if so, go onto the next scheduled task or hold in place till it receives new orders.

So at its heart, we have agents that are situationally aware and able to plan and react according to doctrine. They sometimes react in an uncontrolled manner - eg when they rout, because we model not just hard factors but soft factors like morale, cohesion, and fatigue. The doctrine are a set of rules written as C++ code. So it's a lot of if, then, else type structures. There is a set of plan and task doctrine for each type of task/order. In addition, we have code that handles force allocations, scheduling, intel, pathing, formation movement, resupply and all the base mechanics like firing, bombarding, construction, demolition, deployment, facing, fatigue, cohesion, sighting etc.
 

共工熙雲

Member
Joined
Oct 30, 2017
Messages
70
Points
8
Age
35
Location
China Chongqing
Website
user.qzone.qq.com
The game itself has a "regiment" as a unit of the composite attack system.In the game almost no use. We are almost all to "camp" as a unit to control.In fact AI's "random response" ability is too bad.

If one of the troops breaks through the enemy, and the other side doesn't break, my troops don't go to the other side to support the friendly forces.

There's another question about the enemy's retreat. Fighting late,You will encounter enemies that were scattered before. ,you will be shot before the enemy attack from the rear, frequent harassment, unbearable. This is not possible in reality.

There is sometimes a bug, the troops can not enter the forest situation.
 

Dave 'Arjuna' O'Connor

Panther Games Designer
Joined
Jul 31, 2014
Messages
3,415
Points
113
Location
Canberra, Australia
Website
www.panthergames.com
The game itself has a "regiment" as a unit of the composite attack system.In the game almost no use.
Are you referring to Regiment Headquarters?

We are almost all to "camp" as a unit to control.
Do you mean you have to garrison an objective?

In fact AI's "random response" ability is too bad.
Random response to what?

If one of the troops breaks through the enemy, and the other side doesn't break, my troops don't go to the other side to support the friendly forces.
Show us a screen shot or a save where this occurs. Then I'll be able to see what's going on.

There's another question about the enemy's retreat. Fighting late,You will encounter enemies that were scattered before. ,you will be shot before the enemy attack from the rear, frequent harassment, unbearable. This is not possible in reality.
Oh yes it is. WW2 histories are replete with cases. When a force penetrates and advances deep into enemy lines odds on there will be enemy stragglers, broken units now recovered, enemy units cut off by the initial assault and now trying desperately to find the new front line and reunite with their forces. As an attacker, you have to accept that as you advance you have to devote an ever increasing proportion of your force to mop up stragglers, garrison the flanks of your penetration and your line of communications. That is proper and realistic operational planning.

There is sometimes a bug, the troops can not enter the forest situation.
In most scenarios, motorised units cannot enter forests.
 

共工熙雲

Member
Joined
Oct 30, 2017
Messages
70
Points
8
Age
35
Location
China Chongqing
Website
user.qzone.qq.com
Are you referring to Regiment Headquarters?

Do you mean you have to garrison an objective?

Random response to what?

Show us a screen shot or a save where this occurs. Then I'll be able to see what's going on.


Oh yes it is. WW2 histories are replete with cases. When a force penetrates and advances deep into enemy lines odds on there will be enemy stragglers, broken units now recovered, enemy units cut off by the initial assault and now trying desperately to find the new front line and reunite with their forces. As an attacker, you have to accept that as you advance you have to devote an ever increasing proportion of your force to mop up stragglers, garrison the flanks of your penetration and your line of communications. That is proper and realistic operational planning.


In most scenarios, motorised units cannot enter forests.
Sorry, my English is not very good.
and....
1.Players are almost always "Bn" as a unit to issue orders, when the order of the object issued as "Regt" or more advanced units, AI developed a very scrawled plan.
3.AI is not so smart, they will not respond to the actual situation, but only a simple implementation of the order. Of course, I can say too much.
4.Breakthrough line of defense will not support flank, leading to alone.
like this:
141517doxidiriz3tvp8to.png

6.No, I met the infantry can not enter the situation.
 

Bie

Member
Joined
Feb 22, 2017
Messages
350
Points
28
Age
41
Location
Belgium
As far as I'm concerned I think the AI does a decent job of handling most situations. Haven't really seen it do to much bonehead moves in my playthroughs.

Also, in reply to the AI not responding to certain situations: the fact that units don't seem to react appropriately can be because of orders delay and not necessarily because of bad AI. I've had lots of actions go wrong because I lost the initiative and had to keep replanning them. Planning for anything bigger than a battalion will have so much moving parts that anything unanticipated will cause the execution to go a very differently than expected.
 
Top