Learning bots would consume more processing power, which is at a premium when designing games to run on ageing home computers. Far less processing power needed to create much simpler bots that just obey a series of rules.

For first-person shooter games in particular, the advent of multiplayer online gaming has greatly diminished the need for skilled bots. If you want a challenge against a competent enemy, you play multiplayer, not bots. If you are more interested in single player, then you are probably there for the story and experience, and increasingly hard bots are not really necessary (or even get in the way of you proceeding through the story), beyond having a few levels of difficulty which are achieved by giving the bots more powerful weapons and more health.

"How do you model the learning of perception ... etc"

Have sex. Have babies. Raise children. Be a human.

Make love, not war.

That's deep Anthony.

Hi Michelle, thanks for the reply.

Perception isn't simply a reaction time issue. Consider the problem of a bot perceiving another player standing in front of them. Maybe you would implement a reaction time and action time delay of (making numbers up) 0.4 seconds before the bot could notice, aim at and fire at the player.

But how does that reaction time for perceiving and reacting to another player change with these conditions:
1. All other encounterable players are enemies
2. Some players are enemies, some are friends, its bad to shoot friends
3. The player is 3m away, in daylight, no obstacles
4. 50m away, dim light, partially obscured by foliage, not moving
5. 50m away, dim light, partially obscured by foliage, moving rapidly
6. 50m away, dim light, partially obscured by foliage, moving rapidly, but there are 10 friendlies also in the vicinity also moving about rapidly.
7. 3m away, lying prone, motionless, wearing a ghillie suit
8. all other encounterable players are enemies, the other player is 700m away, but moving rapidly on an open plain.

Perception under the almost limitless circumstances under which a player might encounter another player cannot be boiled down to a single reaction time, and although we have a "good understanding" of how reaction times might change, I don't think we have complex models to determine appropriate reaction times under all these circumstances on the fly for each bot in a game.
Added to the perception variability is the action/personality/circumstances variability. Is it preferable to aim and fire as quickly as possible, sacrificing accuracy, ammunition, stealth and possibly a friendly in order to hopefully take out an enemy before they shoot you? Or is it better to delay action slightly, to give your cognitive systems time to aim, to verify that the target is hostile, to conserve ammo, or avoid giving away your position? Is it better to shoot quickly from the hip at the body of your opponent, or take more time and go for trickier headshot? And what about modifying mid-game? e.g. if the bot shoots too quickly and accidently injures a friendly, will they take more time with their next shot?

These issues/challenges are absolutely bread and butter of playing an FPS for a person, but for a bot? Is this just boiled down to a set time for perception, an set time for aiming, a set parameter controlling accuracy, with a random number generator introducing some realism?

The point I was trying to make was that making a bot that is satisfying to play against isn't just about good strategic/tactical thinking, so that the bot exploits map terrain and behaviour. A satisfying bot would ideally exhibit the same big variety in capacity to perceive other players based on the exact perceptual circumstances.

Part of my own frustration in playing against bots in the single-player FPS out there was not just that the bots seemed to act stupidly, but also that they perceived stupidly. e.g. I might find myself standing next to a bit shooting, but because i wasn't in its field of view, it didn't turn to the sound, or think to look around. Or I might be crouching motionless in foliage in the dark in a game, conditions under which a human player would have difficulty seeing me, but a bot facing me would just start headshotting me with no evident difficulty in seeing me.

While I think improving the tactical/strategic play of bots is very welcome, I'm just wondering whether there is a point beyond which it is no longer useful to do so, because the biggest downside of the bot is no longer its strategy, but its lack of realistic perception.

"militarisation of leisure time"

hyperbole. Its a game involving tactics, strategy and reflexes. People enjoy such games.

Does "militarisation of leisure time" also apply to chess?

You can put different themes or "skins" on chess. Chess can be played with traditional pieces, or realistic medieval pieces or The SimpsonsTM chess, or little soldier and tank figurines or on a computer with animated pieces that attack each other, whatever theme you want. Same with FPS.

You could make an FPS with teddy bears running around throwing love hearts at each other, or giving hugs when up close.

You could make an FPS with kids throwing water balloons.

The popularity of violent "skins" in media including games is much broader than this research, and researching strategy/tactics/AI while obviously relevant to the military, is not exclusive to the military, and shouldn't be avoided just because it might be relevant to the military.

Personally, I find video games boring beyond all belief, and I am fed up with real wars.

Perhaps there could be a video game on how to disarm your enemy.

Or, a video game on how to convince your enemy to disarm.

Hi Dale,

I learned the basic rules of chess when I was about 7. Shortly afterwards, I decided chess was boring, and avoided it for over 20 years. Then, when I was about 30, a flatmate bought chessmaster software, that included many hours of tutorials. We started playing chess against each other, and doing the tutorials to learn better strategy and tactics. At that point, I realised that when I had known the basic rules, and had played a bit, and had concluded chess was boring, I didn't actually know how to play. There is so much more to chess than its superficial rules.

The same applies to "video games".

For starters, the types of video game out there, the themes, play styles, depth, multiplayer vs single player vs cooperative play etc are so diverse, that to say something like "video games are boring beyond belief" makes about as much sense as saying "reading is boring beyond belief" because you read and were bored by a copy of New Idea.

Should war break out between China and the US, we may be thankful for frontline troops wearing exoskeletons ~ and the exoskeletons would in fact be robots.

The skills of the human and the machine would then merge and network across a grid, supported by packs of robodogs and drone-bot swarms for each exodigger.

We may then be thankful that gaming skills were developed that enabled young soldiers to defend our nation.

The training of exodiggers may happen in the virtual worlds, as amazingly, robots must learn and can learn ~

http://modemworld.wordpress.com/2011/09/24/moses-the-us-armys-opensim-exercise/

This skill of war is already finding its way into the health services, enabling the disabled to walk ~

http://www.fastcoexist.com/1680940/this-light-affordable-exoskeleton-could-help-the-paralyzed-walk-again#1

Soon the disabled will be running and even paraplegics helping themselves to the cookie jar ~

http://www.abc.net.au/news/2012-12-18/paralysed-woman-controls-robot-arm-with-brain/4432994

For some folk the game of life is quite tiring, but through gaming and robotics, a new world of opportunity is being born.

Kim Peart

Anthony,

I'm not entirely sure what your point is or why you've chosen what is an interesting article on the Bots in FPS games to try and make it. There are both military and non-military FPS games, nobody is forcing the "militarisation of leisure time".

I've seen a few of your comments on similar matters and they often remind me of a piece by Dave Grossman called "On Sheep, Wolves and Sheepdogs".

Dale Bloom ~

In my document ~ 'Creating A Solar Civilization' ~ found on my website ~ I explore how the way to achieve peace on Earth and send war into history, is through expansion beyond Earth.

I found that there were feedback loops that come into play, once we begin expanding into the celestial realm.

To achieve this future and therefore peace on Earth, will require considerable effort and challenge our imagination to the hilt.

We will need to inspire a critical number of people on Earth to rise to the challenge, to change our game from an Earth-trapped carbon economy to a space liberated stellar civilization.

With a need to drive a global campaign that any number of people could be directly involved in ~ even on a daily basis ~ I came to see the benefit of a form of Internet gaming, called the virtual world, where participants make the content.

The virtual world doesn't seem like a game, but games can happen in it, as do robots, which learn through engagement.

If my observations prove to be correct, then the way beyond war lies through gaming and celestial expansion.

Without the gaming world and a celestial future, we appear to be locked into endless wars on Earth, as climate change delivers an increasing depth of terror.

I hope we wake up and choose to play the celestial option, rather than running blindly into hell on Earth.

Kim Peart
http://www.islandearth.com.au/

James Wookey ~

An excellent article, that 'On Sheep, Wolves and Sheepdogs' by Dave Grossman. ~ The Conversation is quite an education.

Reference to "honour" for Australia must be in the context of the great power we follow, as a vassal state.

Our "moment of truth" arrived in 1962 and we lost it.

There was a simple honour in preparing the Papuans of western New Guinea for independence and we had been working with the Dutch since 1957 to achieve this, with many of our people on the ground in West Papua, raising West Papuan hopes of self-determination.

When Indonesia began invading New Guinea in 1961, the US decided that friendship with Jakarta was more valid than Papuan freedom.

We did what we were told, walked away from the Papuans and allowed the slow-motion genocide to proceed under Indonesia, first in West Papua, then in East Timor and still continuing in western New Guinea.

For Washington, the selling of the Papuans to Indonesia was a necessary slave trade on the road into Vietnam, where 7 million tons of bombs would be dumped to send a couple of million people into oblivion.

Vietnamese children still die from unexploded bombs.

We wouldn't fight the war we could have won in New Guinea and lost the war we dreamed of winning in Vietnam.

Will we ever regain our moment of truth and honour with West Papua?

We might, if we lift our game considerably, which I explored in this article ~
http://www.independentaustralia.net/2011/environment/avoiding-calamity-by-looking-to-the-stars/

Undoubtedly, politics is a dirty game, when it swipes away our honour and truth.

If we can find our heart and our honesty, then we may find our way beyond the dirty games of realpolitik that drag us into the slime of ignominy.

Kim Peart

Like Kate, I am a RPG kinda guy. These are specifically tabletop RPGs (akin to DnD, Call of Cthulhu etc) but I do play MMOs when I need a distraction from real life.

I've recently started mucking around with a set of game engine software, specifically because I am getting bored with the robotic nature of most NPC AIs.

For instance, in Fallout3, I got bored with the kamikaze nature of some of the AIs, when it came to combat. One after another they would just charge in. No sneaking about, no tactical withdrawal, no leaving a PC alone as they were too tough to crack.

The potential for AIs to start stupid, and slowly learn from interaction with players is, to be honest, beguiling. Work like yours is potentially of great interest to me.

AIs that act intelligently, are not predictable, and can adapt their in-game interaction with both the game world and the players.... now *that* would make for interesting gameplay.