Table of Contents

Quick Start...4
SYSTEM REQUIREMENTS..4 INSTALLATION.. 4 MAIN MENU...5 MULTIPLAYER GAME..7 GAME CONTROLS...10
Operation Blintz...11 Getting Started...12
HEADS UP DISPLAY..12 BASIC STRATEGY...12 RESOURCES...14 MAP...15 UPGRADES...15 SQUADS...16 RALLY POINTS (FLAGS).. 16 SARGES OUTFIT..16 RELATIVE DAMAGE/STRENGTH..19
2002 The 3DO Company. All Rights Reserved. 3DO, Army Men, Real Combat. Plastic Men., and their respective logos,are trademarks or registered trademarks of The 3DO Company in the U.S.and other countries. GameSpy and the GameSpy logo are trademarks of GameSpy Industries,Inc. All other trademarks belong to their respective owners. The software and related manual for this 3DO product are copyrighted. No portion of this product may be photocopied, scanned, translated, reproduced, copied or reduced to any tangible or electronic medium or machine-readable form, or publicly performed or displayed, without the prior written consent of The 3DO Company. The owner of this product is entitled to use the software and related manual for his or her own use,but is not entitled to reproduce and distribute any copies of the software or manual to any other individual or entity; nor to rent or lease this product or any copy thereof to any third party. Windows 95,Windows 98,DirectPlay and DirectX are trademarks of the Microsoft Corporation.
Components..20 Power-Ups...27 Secret Codes...27 Pause Menu: Saving and Loading.28 Customer Support...29 Limited 90-Day Warranty...30

Quick Start

System Requirements
Make sure that your computer meets all of these requirements before installation. Operating System: Windows 95/98/Me/2k/XP home CPU: Pentium 233 (PII400 recommended) Free Hard Disk Space: 250MB Memory: 64MB RAM minimum (96MB recommended) CD-ROM Drive: 4X Speed Video Card: DirectX 8.0 compatible at least 16MB VRAM (32MB recommended) Sound Card: All major DirectX 8.0 compatible sound cards Peripherals: Keyboard and mouse Multiplayer: 56k modem or faster connection to the Internet. TCP/IP protocol for the Internet or LAN play Note: May not run from CD-RW and DVD drives. You must have Winsock v2.0 (comes installed in Windows 98, 2000, Me and XP home) and DirectX 8 installed on your computer. Both of these files are available directly from http://www.microsoft.com.

Main Menu

BOOT CAMP (TUTORIAL)
Boot Camp offers three lessons in real-time strategy basics: 1.Camera and Movement: Learn to set the camera position and how to command each Soldier. 2.Building a Base: This lesson is essential for knowing how to build and set up camp, upgrade each building, and create your army. 3.Squads and Combat: The final lesson of the tutorial teaches you how to create individual squads and engage the enemy.

CAMPAIGN

Campaign mode takes you on 15 action-packed missions through the backyard and into the house in your search for the renegade soldier, Blintz.

GREAT BATTLES

Installation
1. Place the CD-ROM labeled Army Men: RTS into your CD-ROM drive and follow the onscreen instructions. If your CD-ROM does not autorun, double-click on the CD-ROM drive icon under My Computer from your desktop. 2. When you see a listing of the files on the CD-ROM, look for AUTORUN.EXE and double-click to install Army Men: RTS. 3. Press the Install button to begin installing and follow the onscreen instructions. 4. View the Readme file that contains last minute information and changes that could not be printed in this manual. If you have any problems with Army Men: RTS, please review the Readme file for any known problems and workarounds.
If youre in the mood for simple short skirmishes, Great Battles allows you to choose from a list of unique single-objective missions. Try missions like eliminating all Tan structures, blowing up the Tan base, or clearing the bathroom towel rack of all Tan troops! Great Battles are unlocked as you earn Gold Medals in the campaign.

LOAD GAME

From here you can load a previously saved game.

MULTIPLAYER

Go here to start or join a Multiplayer Game. (see Multiplayer Game section on page 7 for more specifics)

OPTIONS

The Options Menu is separated into the following sections:

VIDEO OPTIONS

Multiplayer Game
Now you can challenge friends to tactical battles online, either on a Local Area Network or over the Internet. As many as eight people can play. Join up as teammates or play head-to-head. The goal is simple: defeat your opponents troops and destroy his or her Headquarters. If you are able to destroy the Headquarters and your opponent is unable to rebuild it within 3 minutes, you win! Likewise, if your Headquarters is destroyed and cant be rebuilt in time, youre out of the game.
FOR A MULTIPLAYER GAME OVER A LOCAL AREA NETWORK:

Screen Resolution and Color Depth Here you choose among seven different options. High performance computer systems can run at higher resolutions than low performance systems. If you find the game running slowly, you can modify this setting to a lower resolution. World Detail & Texture Detail Again, higher performance computers should use a higher setting.

AUDIO OPTIONS

Select MULTIPLAYER from the Main Menu. From here you can create a game, select an existing game, create or change your player name and access your address book for additions or modifications. Select Change Name to create your player name. Enter the name you wish to give yourself and select OK. If someone is already hosting a game, the game name will appear in the dialog box; select it if you want to join in. To Host a new game, select Create Game, enter a name for your new game and select OK. Now you can choose from several multiplayer missions by clicking on the Select Mission button, highlighting a mission and selecting OK. The current mission map is displayed beneath the Select Mission button. You can choose to let the computer control a team by selecting the Add AI Player button. Wait for other players to join in (their names will appear in the Players Box). Select the Lock Game button if you dont want anyone other than the players listed to join in. Select a color for your troops by clicking on a color sphere. The selected color will appear next to your player name in the Players Box. Once you have made all the modifications you want, and all the players names appear in the Players Box, select Play to start the mission.
You can set the Master Volume and the Music Volume separately with these controls.

DIFFICULTY

We recommend playing the game on Normal difficulty. However, some missions may require playing on Hard difficulty to be awarded a gold medal.

CREDITS

Select this option to view the game credits.
Select this to quit the game and return to the desktop.
HOW TO PLAY ARMY MEN: RTS IN GAMESPY ARCADE:
FOR A MULTIPLAYER GAME OVER THE INTERNET:
You can play Army Men: RTS online through GameSpy Arcade, which comes conveniently bundled with the game. If you haven't done so already, insert your Army Men: RTS CD and install Arcade now. Then, to play Army Men: RTS online, just follow these simple instructions: Launch GameSpy Arcade and Go to the Army Men: RTS Room: Click on the GameSpy Arcade link in your Start Menu. When the software starts, you'll see a list of games and more along the left-hand side. There's plenty to do, but there's time for that later. Click on the Army Men: RTS button on the left to enter the Army Men: RTS room. Find or Start an Army Men: RTS Server: Once you're in the Army Men: RTS room you can meet or greet other players, find servers or create your own server. The top half of the application will list all of the available servers, including the number of people playing and your connection speed measured by something called "ping." The lower your ping, the better. Don't like any servers? Click on the "Create Room" button to start your own server and wait for people to sign up. (A clever server name, such as "Taste the end of my rusty hob-nailed boot!" usually attracts people.) Otherwise, double-click on a server of your choice to join in. Joining and Starting a Game: Once you double-click on a server or start your own, you'll be in a staging room, in which you can trash talk with your fellow players and prepare for combat. When you're ready to play, click the "Ready" button at the top of the screen. When everyone in the room has signaled their readiness, the host can then launch the game. Arcade will fire up Army Men: RTS and the carnage will begin! Problems? If you have problems using Arcade, whether installing the program, registering it, or using it in conjunction with Army Men: RTS, consult our help pages, located at http://www.gamespyarcade.com/help/ or e-mail us by using the form located at: http://www.gamespyarcade.com/support/contact.shtml.

Select MULTIPLAYER from the main menu. From here you can create a game, select an existing game, create or change your player name and access your address book for additions or modifications. Select Change Name to create your player name. Enter the name you wish to give yourself and select OK. If you want to join an existing game, select Address Book and select the IP address of the host of the game you wish to join. If the address has not yet been entered, select Address Book, and select Add. Enter the name and IP address of the person hosting the game, and select OK to add it. You can edit any address at any time by selecting Edit on the Multiplayer screen. To Host a new game, select Create Game, enter a name for your new game and select OK. Now you can choose from several multiplayer missions by clicking on the Select Mission button, highlighting a mission and selecting OK. The current mission map is displayed beneath the Select Mission button. You can choose to let the computer control a team by selecting the Add AI Player button. Wait for other players to join in (their names will appear in the Players Box). Select the Lock Game button if you dont want anyone other than the players listed to join in. Select a color for your troops by clicking on a color sphere. The selected color will appear next to your player name in the Players Box. Once you have made all the modifications you want, and all the players names appear in the Players Box, select Play to start the mission.

Operation Blintz

Game Controls

FUNCTION MOUSE

Select Unit. Left Mouse Button Select Multiple Units. Left Mouse Drag Move Selected Unit(s) to. Left Mouse Button Location Selected/Attack Deselect Unit. Right Mouse Button Change direction of building. Middle Mouse Button, before construction "<" and ">" Keys Scroll Camera Around. Right Mouse Drag Camera Height. Mouse Wheel

FUNCTION KEYBOARD

Create squad with currently. Ctrl + [number key] selected units Select Squad. [number key] Bring up Game Menu. ESC Move camera to the last point of combat. Spacebar Stop Units. s Center Camera Over Headquarters. h Camera Height. F1 F4 Quick Save. F9 Save/Load Game. F10 Scroll Screen. Arrow Keys
Ive seen plenty of action in my time. Ive led troops into countless skirmishes, situations where we knew wed either make it back to fight another day or be reduced to puddles of plastic on the battlefield. But weve never had to deal with anything like this before. This mission is bad. Worst I've ever seen. Not long ago, we had this region secure under the supervision of one of our own, Colonel Blintz. That changed when Blintz was injured in the line of duty. Medics did what they could, but he sustained a "massive, disfiguring head wound". Long story short, theres a chunk of plastic missing from the top of his head. "Disfiguring head wound"? Hes lost his mindliterally! That was the last we had heard of him. Thats when our reconnaissance got on the case and foundwell, it aint pretty. With the help of a bucket of paint, Blintz has gone Tan. Our mission is to find Blintz andterminate him. This will be the toughest mission Ive ever had to lead. Blintz was famous for inspiring loyalty. His men would march across a grill if he ordered it. And there are a lot of them. This wont be easy. - Sarge

Getting Started

Listen up, Soldierthis is a war weve got on our hands! If you want to make it to the end of the day, youve got to know this platoon inside and out.

Heads Up Display

2) Use a Bulldozer to build a Resource Depot reasonably close to any resources you see, and then send your Dump Truck(s) to the resource to collect it. Follow the instructions listed in Step 1 to build the Resource Depot. 3) When your HQ is built, youre ready for recruits! Use the Bulldozer to build a Barracks and create a basic squad comprised of Grunts, Grenadiers and a couple of Mine Sweepers. With your squad ready for action, send them out to explore the area immediately around your base. Follow the instructions listed in Step 1 to build the Barracks. To create Soldiers, select the Barracks. The Construction Menu with Soldier icons will appear on the left. Select the Soldiers you would like to create. Every click on a Soldier icon is another Soldier to be made. The total number of Soldiers being created appears over the icon. If you change your mind, simply right-click on the Soldier icon to subtract from the queue. Set a Rally Point for the Soldiers. (See the "Rally Points" (Flags) section in this chapter.) Select the group of Soldiers you want to explore the area with by left-clicking, dragging a box around them, and releasing the left mouse button. Left-click on the point of the playfield you would like them to explore. Note: When a Soldier type is highlighted in the Construction Menu, his relative strength vs. Infantry, Armor and Air enemies is displayed on the right side of the screen. This info is useful when you are deciding which soldiers you will need to fight against the enemies you currently face. 4) Once youve sent your squads out to explore the area, youll have a general idea of where the Tans attack from and the basic layout of the land. With these things in mind, build a Barbed Wire Fence around your camp. Its also a good idea to put a few Guard Towers between pieces of fence. The Tower functions as a lookout and has the firepower of several Grunts.

PLASTIC

ELECTRICITY MELT BUTTON

MAP CONSTRUCTION MENUS

SQUAD MENU

Basic Strategy

There are several steps you will need to follow in most missions. 1) Begin base-building missions by building a Headquarters (HQ). This allows you to build several Dump Trucks. In addition, the HQ is a prerequisite for all buildings except the Resource Depot, so youll want to have one at all times. You can also build Bulldozers from the HQ should you lose one to the enemy or want more than one. To build a structure, select the Bulldozer and your choices appear on the left in the Construction Menu. Left-click on the structure you would like to build. A ghosted icon of your new structure appears on your cursor. Move the cursor over the place on the playfield you would like to build it. (If the image is red, it cannot be placed in the current location.) Use the "<" and ">" keys to rotate the structure to the desired position, and then left-click to build it.
5) When youve got the resource reserves built up enough, upgrade your HQ to a Super HQ and build a Garage. Use the Garage to build Medic jeeps that can heal your troops, your buildings and each other. You can also create Half-Tracks (aka: machine gun trucks), Minelayers and Tanks. See the "Upgrades" section of this chapter. Follow the instructions listed in Step 1 to build the Garage.

COLLECTING

in more advanced missions. 6) When you have the resources, upgrade your Barracks to a Super Barracks. You can use this new structure to pump out a more specialized platoon: Bazooka Men, Machine Gunners, Snipers and Mortar Men. 7) Finally, when you have the resources available, upgrade the Garage to the Super Garage and make DumDums and Choppers for assaults on the Tan base! 8) Further upgrades will allow you to take on the enemy even more effectively. NOTE: If any icon in the Construction Menu appears red, you either dont have the necessary requirements or the resources available to build it yet.
Use Dump Trucks to collect all Plastic and Electricity. Water guns, dog bowls, batteries, and walkie-talkies are just a few of the items you may find. When you locate a resource, build your Resource Depot nearby. Your trucks will collect resources faster if they only have to move a few feet to deliver the goods. Be sure to check your resources from time to time to see if theyre being depleted. The amount of collected Plastic and Electricity appears in the upper-right of the game screen. If a resource runs dry, move the Dump Trucks to the next available resource and continue the hunt. It is also helpful while one Dump Truck works on collecting resources, to have another one looking to collect melted plastic Soldiers, vehicles, etc. Dump Trucks collect resources automatically, and do not need to be directed. However, you can override the automatic function by manually assigning them to a resource: To manually send a Dump Truck to mine a resource, left-click the Dump Truck to select it and left-click the resource you want it to collect.

MELTING STUFF

Resources
Resources are the vital components that allow you to create items and build up your army. In order to survive, you need to keep building, and in order to keep building, you need to keep collecting resources: Plastic and Electricity. While you may not have an immediate need for resources, its always a good idea to keep collecting and storing so you have them when you need them. Be sure to keep an eye on each resource as its being collected, as each will eventually become depleted.
If a structure becomes unnecessary, melt it down and youll get half of your resources back. To melt a structure down, simply left-click it to select it. The "Melt" button will appear in the upper-right corner of the screen. Select it to melt the structure.
The map is your most useful tool for keeping apprised of everything at once; use it to immediately jump to enemy units or your own squads. When you get a red warning that your units are under attack, double-click on the map or press the Spacebar to instantly jump to the action!

Upgrades

Whenever you have collected enough Plastic and have reasonable defenses in place, you can begin upgrading your buildings. Youll want to have the most specialized equipment before the Tan do! The buildings that can be upgraded are: the HQs , the Barracks and the Garage. You can upgrade each one on its Construction Menu.

Squads

You can form up to four unique squads. This makes independent control of each group on multiple battlefronts much easier! Highlight a group of Soldiers and press Ctrl + 1, 2, 3, or 4 to assign a number to each group or click on a Squad icon at the bottom of the screen. To instantly highlight a squad and give them a command, press the corresponding key: 1, 2, 3 or 4, or select the Squad icon at the bottom of the screen.
To select a Soldier and move him, left-click on the Soldier and then on the place on the playfield you want him to move to. If an enemy is sighted, the Soldier automatically launches an attack and radios a message to you. To select a group of Soldiers, left-click, drag a box around them, and release the left mouse button. Left-click on the point of the playfield you would like them to move to.

SERGEANT HAWK

Rally Points (Flags)
As you build troops and vehicles, you can set up Rally Points to immediately ship your units to any specified location the moment they pop out of the mold. This automates your troops, allowing you to concentrate on the mission at hand. Highlight the structure (either the Barracks or the Garage) by pressing the left-mouse button. Left-click anywhere on the playfield to set the Rally Point, a Flag icon will appear. (Right-click will cancel Flag placement.)
Sergeant Hawk, a.k.a. Sarge, is the Bravo Company Commandos unquestioned leader. Whether the battle is in a town, on a bridge or in a backyard, Sarge is itching for the chance to send the Tan Army back to the plastic goo from which they were molded! Sarge has plenty of motivation to fight the Tan menace - he must save his friends in Bravo Company to keep the Green Nation safe and to protect the woman of his dreams, Vikki.

Vikki is Colonel Grimms feisty and lovely daughter. She has a spirit for adventure and has chosen the occasionally dangerous occupation of reporter. Shes able to take care of herself and is practically one of the Bravo Company Commandos - only her beauty singles her out from the rest of the war dogs! Sarge is the guy for her.

Sarges Outfit

As you advance through the missions, you may recognize these faces: Sarge, Hoover, Scorch, and many other members of Bravo Company. Some or all of these fighters are the first to appear in each mission. Assign them to the defense of your base until you get the Barracks up and running, then lead them into battle! They are leaner, meaner and tougher than any Soldiers you can create and have more hit points, so send them out with your squads and into battle. Use them wisely!
Riff is itching to play some tunes on the Tan Army in the key of B, for Bazooka! Riff is the heavy hitter of Bravo Company and the Commando the Green Army needs when serious thumping on the Tan Army is required.

HOOVER

Private Hoover has got the most stressful job in Bravo Company and it shows, judging by his easily startled nature. Still, he is a heroic soldier who is called upon whenever a mine field needs sweeping.

Relative Damage/Strength

Shrap is a Mortar Man extraordinaire. Although he would rather be catching a wave than riding the sound wave of a large mortar explosion, Shrap is happy as long as hes dropping some heinous mortar shells on the Tan Army.
Thick is about as sharp as a marble. What the big lug lacks in brains, he makes up for with brawn and enough cover fire with his M-60 machine gun to knit a lead blanket. Sarge has a soft spot for Thick and the enthusiasm he brings to the Bravo Company Commandos.

RELATIVE DAMAGE

The chart above shows how much damage each component can take.

SCORCH

Scorch likes fire just a little too much, considering that hes made of plastic. Theres nothing Scorch loves to hear more than the sound of Tan Soldiers bubbling after he has laid on some serious heat with his flamethrower.

BULLSEYE

Bullseye, the newest Commando in the company, has a keen eye. His sharp shooting can drop a Tan Soldier in his tracks even before the others know danger is coming. He keeps his cool in any situation - the ice man of the Bravo Company Commandos.

RELATIVE STRENGTH

The chart above shows how much damage each component in the game causes the other components.

Components

Everything revolves around Plastic and Electricity the more resources, the bigger the army you can create. The mission itself determines the amount of Plastic you begin with.

GUARD TOWER

Buildings

HEADQUARTERS (HQ)

Plastic required: 200 Guard Towers are very effective for providing defense around the perimeter of the base and at the main entrance, as well as against Tan air attacks. Guard Towers can link to Barbed Wire Posts.
Plastic required: 400 This is the most basic component in your base. The HQ allows you to build Bulldozers and Dump Trucks that are necessary for building every other unit in the army. This should be the focal point of your base.

RESOURCE DEPOT

GARAGE
Plastic required: 400 The Garage gives you the ability to build heavy assault vehicles and one of your most important assets, the Medic jeep.
Plastic required: 500 The Resource Depot is the structure in which you process all of the Plastic and Electricity the Dump Trucks collect. Guard the Resource Depot well!

BARRACKS

PILLBOX
Plastic required: 300 This is a small building which allows the Soldiers inside to be protected, yet have the ability to shoot 360 from a single position. They are vulnerable from the air.
Plastic required: 300 The Barracks creates the bread and butter of your army: your infantry. Your Barracks is your first line of attack. and defense. Units built from the Barracks do not require electricity, which can be scarce at times.

BARBED WIRE POST

ANTI-AIRCRAFT GUN
Plastic required: 150 This stationary gun mount for defense against enemy Choppers is a must for every army base.

Plastic required: 50 Use Barbed Wire Posts to build fences to hold back the enemy and keep them out of your base. It is cheap and fast to build.

Soldiers

BAZOOKA MAN

Plastic required: 125

Plastic required: 50
You can never have enough Grunts in your platoon. They operate by targeting one thing and firing at it continuously with rifles. They are not only the cheapest unit you can create, but are also effective against the entire Tan Army.
These plastic Soldiers use their bazookas for long-range attacks, and are very effective against Half-Tracks, Tanks, Choppers, and buildings.

SNIPER

Plastic required: 250

GRENADIER

Plastic required: 75
While a Snipers rate of fire is fairly slow, his gun has the longest range of any Soldiers. However, they are only effective against personnel.
Grenadiers are the other staple to every squad. These highly effective fighters throw grenades for an area-effect attack and are cheap to produce, so dont be afraid to populate the platoon with them. However, they cannot attack air vehicles.

MORTAR MAN

Plastic required: 150

MINE SWEEPER

Mortar Men attack structures only, so send them out directly to each Tan base for some serious long-range wreckage. Remember, if they can see it, they can destroy it.
While Mine Sweepers are not fighters, they are very useful for detecting Tan mines.

RADIO OPERATOR

Plastic required: 400

MACHINE GUNNER

Plastic required: 100
Like Bazooka Men, Machine Gunners have the ability to fire long-range guns. Also, their guns shoot much more rapidly than Grunts fire.
Bring in a Radio Operator for some incredible attacks! Use Radio Operators to call in: Paratroopers A giant Magnifying Glass for massive Tan melting Bombing Runs

Structure Upgrades

SUPER HQ

Vehicles

BULLDOZER

Plastic required: 200

Upgrade the HQ to a Super HQ to increase your build options.

SUPER BARRACKS

Upgrade your Barracks to get the Super Barracks. This new structure gives you the ability to produce the best of the best fighters: Bazooka Men, Machine Gunners, Snipers and Mortar Men. These Soldiers have increased range and more powerful guns than the basic Grunts.

SUPER GARAGE

Bulldozers are essential to building all camp structures: Barracks, Guard Towers, HQ, Resource Depots, Barbed Wire Posts, Anti-aircraft Guns, Garages and Pillboxes. Keep them safe from the enemy as they are key to getting your base up and running, and keeping it active.

DUMP TRUCK

Dump Trucks are the key to collecting resources. They gather the Plastic and Electricity needed to build your army.

MINELAYER

When the Plastic is available, upgrade your Garage to a Super Garage. This will give you access to the next level of equipment, including Choppers!

SUPER-DUPER HQ

Plastic required: 100 Electricity required: 100
The Minelayer is useful in situations where your army is on the move and you need to protect your flank and lay mines.

Plastic required: 600

This will let you build a Radio Operator. Radio Men will let you build Paratroopers, a Magnifying Glass and Bombing Runs.
Plastic required: 100 Electricity required: 150
A Medic jeep allows you to heal all of your wounded Soldiers, any damaged structures and even other damaged Medics. 25

POWER-UPS

Vehicles cont.

HALF-TRACK WEAPON BOOST

This power-up increases the firepower of all your units.
The Half-Track is a good all-around assault vehicle, and its ability to hit air targets, as well as ground troops, makes it a useful addition to any squad.

SPEED BOOST

This power-up increases the rate of movement of your troops.

HEALTH BOOST

This power-up gives full health to all of your units on the map.
Plastic required: 200 Electricity required: 250
Tanks are well-known as powerhouses in any assault. Just watch out for Choppers. These guys cant do anything but run away from them.

SECRET CODES

For destroying specific objects in some missions, the player will be awarded a secret code on the Victory screen. (When the object is destroyed, a sound effect will be heard). Entering the secret code: While in any mission, hold down the ALT key and press the Backspace key. This will bring up a dialog box. Type in the secret code and press the Enter key. You may enter most codes only once per mission.

CHOPPER

Plastic required: 150 Electricity required: 250
Choppers are great for quick aerial assaults on enemy positions.

DUMDUM

Plastic required: 50 Electricity required: 150
Send these small robots out to the enemy lines where they can deliver the explosives strapped to their backs up-close and personal. 26 27
Pause Menu: Saving and Loading
At any point in the game, simply press the ESC key to pause the game and bring up the Pause Menu. The Pause Menu offers the following options:

RESUME GAME

Customer Support
However you choose to contact us, there are some pieces of information about your computer that we will need in order to help you properly. You can get this information by following these steps: First, right-click on the My Computer icon and choose Properties. On this System Properties screen, you will find the following information that we will need: Computer OS type (Windows version), amount of RAM and CPU type. After you get this information, exit the System Properties screen, then double left-click on the My Computer icon. On this screen, right-click on the icon of the hard drive that the game is installed on and choose Properties. On this screen, you will need to find out the drives total capacity, and the amount of free space left on the drive.

WEBSITE

Select this when you are ready to return to gameplay.

MEDAL GOALS

This displays the Medal Goals for the current mission.

SAVE AND LOAD

Save the game or load a previously saved game.
Adjust your Gameplay or Audio Options.

RESTART MISSION

Visit www.3DO.com/support for late-breaking news and information, answers to Frequently Asked Questions (FAQs), links to hardware manufacturers websites, and the latest product updates. If you want to ask a specific technical question, you can use the online email form available at our website or send e-mail directly to: customer-support@3DO.com.

AUTOMATED PHONE AND FAX

If the mission is going badly, select this option to begin again.

ABORT MISSION

Quit the current mission.
Call (650) 385-3193 or fax us at (650) 385-3181.

U.S.MAIL

Quit the game. NOTE: The game cannot be paused in Multiplayer mode.
Customer Support The 3DO Company 100 Cardinal Way Redwood City, CA 94063

BUY DIRECT

If you wish to purchase additional copies or other titles from The 3DO Company, please visit the 3DO Direct Online Store at http://store.3DO.com, or call: 3DO Direct Sales

WORLD WIDE WEB

(800) 336-3506 in the United States (650) 385-3187 in Canada and Mexico

http://www.3DO.com 28 29

THE 3DO COMPANY LIMITED 90-DAY WARRANTY
The 3DO Company ("3DO") warrants to the original consumer purchaser of this 3DO software product that the media on which the underlying computer program is recorded is free from defects in materials and workmanship under normal, recommended use for ninety (90) days from the date of purchase. The 3DO software program recorded on such a media is sold on an "as is" basis, without any warranty or condition of any kind, and 3DO shall not be liable for any losses or damage of any kind or nature resulting from the use or inability to use such program. 3DO's entire liability and the original consumer purchaser's exclusive remedy for breach of any warranty (express or implied) shall be, at 3DO's option, either: (a) to replace, free of charge, this 3DO software product, or (b) to return to the original consumer purchaser the price paid for this 3DO software product. The preceding warranty shall not be applicable and 3DO shall not be obligated to replace this 3DO software product or to return any such sum to the original purchaser unless this 3DO software product is returned within the 90-day warranty period,postage pre-paid with proof of purchase to The 3DO Company, 100 Cardinal Way, Redwood City, CA 94063, Attn.: Customer Support Dept. The preceding warranty shall be void if any defect in this 3DO software product arises through accident,negligence, use in any application for which this 3DO software product was not designed or intended,modification without the prior consent of 3DO, any cause external to the product (e.g., heat), or by any other cause unrelated to defective materials or workmanship. EXCEPT FOR THE LIMITED WARRANTY SET FORTH ABOVE,ALL OTHER REPRESENTATIONS, WARRANTIES,CONDITIONS,TERMS AND OBLIGATIONS,WHETHER EXPRESSED OR IMPLIED, INCLUDING,BUT NOT LIMITED TO,ANY IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS FOR ANY PARTICULAR PURPOSE AND/OR NON-INFRINGEMENT REGARDING THIS 3DO SOFTWARE PRODUCT (WHETHER RELATING TO THE MEDIA,THE SOFTWARE OR OTHERWISE) ARE EXCLUDED TO THE MAXIMUM EXTENT ALLOWED BY APPLICABLE LAW. 3DO neither assumes nor authorizes any other person or entity to assume for 3DO (or any related or affiliated entity) any other liability in connection with this 3DO software product. IN NO EVENT SHALL 3DO BE LIABLE FOR ANY INCIDENTAL, CONSEQUENTIAL (INCLUDING LOSS OF PROFITS), SPECIAL AND/OR OTHER DAMAGES OF ANY KIND OR NATURE RESULTING FROM THE PURCHASE,POSSESSION,USE, OR INABILITY TO USE THIS 3DO SOFTWARE PRODUCT, EVEN IF 3DO HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES OR LOSS. IN NO EVENT SHALL 3DO'S LIABILITY TO THE ORIGINAL CONSUMER PURCHASER EXCEED THE AMOUNT ACTUALLY PAID FOR THIS 3DO SOFTWARE PRODUCT. SOME STATES/COUNTRIES DO NOT ALLOW LIMITATIONS ON HOW LONG AN IMPLIED WARRANTY LASTS AND/OR THE EXCLUSION OR LIMITATION OF INCIDENTAL OR CONSEQUENTIAL DAMAGES, SO THE ABOVE LIMITATIONS AND/OR EXCLUSIONS OR LIMITATIONS OF LIABILITY MAY NOT APPLY TO YOU. THIS WARRANTY GIVES YOU SPECIFIC LEGAL RIGHTS AND YOU MAY ALSO HAVE OTHER RIGHTS WHICH VARY FROM JURISDICTION TO JURISDICTION. 3DO has endeavored to ensure that the end user manual and promotional materials regarding this 3DO software product accurately reference the product. However, because of ongoing improvements and updating of 3DO software products, 3DO cannot guarantee the accuracy of printed materials after the date of publication, and 3DO disclaims any loss, liability and/or damages, whether in contract, tort or otherwise, arising out of or resulting from the manual and promotional materials,including, without limitation,any loss or liability resulting from changes, errors or omissions with respect to any of such items.

Chapter 1. Real-Time Strategy High-level Planning

Chapter 1

Real-Time Strategy High-level Planning

1.1 Abstract

This paper will give a fast overview of the current real-time strategy (RTS) plan making algorithms used in commercial games and in academic research. It will explain why RTS is an important and hard area to design Articial Intelligence for. It discusses in short the strong and weak points of dynamic scripting, Monte Carlo planning and Case Based Planning. These techniques should eventually lead to an AI system that can outperform experts in this domain. Although that is still far away.

Introduction

Articial Intelligence is being used since the very rst computer games. In the past decade the popularity of computer games has increased and the revenue of the gaming industry exceeds multiple billions of dollars [Buro, 2003]. A certain genre of computer games, real-time strategy (RTS), also increased in popularity. In this type of game, playing against other humans is much more popular than playing against the articial opponents. This is because commercial producers of RTS have not been able to enhance the Articial Intelligence capabilities of their games without spending a large amount of money. Furthermore, the CPU time available for real-time strategy AI is limited in commercial games. [M. Buro, 2003] Designing AI for a real-time strategy game is complex. The domain has several issues that need to be addressed by the AI. These can be found in the next section. Apart from the issues mentioned there, the RTS domain is also complex because of the large amount of objects in the game. Objects are units, buildings or any kind of non-static terrain features, such as trees (they can be cut down). [Danny C. Cheng, 2004] This paper will discuss the commercial techniques and the techniques currently being researched. To give an overview of the plan-making methods used in a real-time strategy AI that is designed to win in a fair way.
The next section will discuss the domain of real-time strategy articial intelligence, how does a RTS game work and what problems are there for AIs to address. The Commercial techniques section is about the only commercial technique used in current real-time-strategy games. The Academic techniques section discusses the techniques to win RTS games that are researched in academic world. In the Discussion section we discuss the importance of research in this area.

Real Time Strategy Games

Real-Time strategy games are computer games in which the player controls an army in real-time to destroy other players armies. A simplied military simulation. [Buro, 2003] Typically an army rst has to be built for which one needs resources and buildings. These buildings will have to be Pixel Articial Intelligence Indepth research 1

Simulations & Games

constructed. To construct stronger units the player needs to do some sort of technology research often available in certain buildings. This research takes time and resources to complete. Furthermore a player normally has only partial information about the world around his base and needs to discover the world as his units traverse it. All units and buildings have a certain line of sight in which the player can see enemy units or buildings. The key to winning a game is to balance the army production with the gathering of resources and technology. Concurrently the player should send scouts to discover enemy bases, armies and resources to destroy them. When all enemy units and buildings have been destroyed the player has won. The tasks in real-time strategy games can be split into three levels of abstraction. Unit control Control at the lowest level, here the player controls the specic unit. For example, he can order a unit to destroy an enemy unit. Or he orders one of his workers to build a tower. Units have a standard behaviour at this level, even for human players. This standard behavior is normally a simple script-based system. With rules like: if enemy in range, attack. Players can increase their strength by overriding the standard behaviour. For example by focusing attacks on individual enemy units to quickly reduce their numbers. [Michael Chung, 2004] Tactical planning Tactical planning consists of making plans on how to attack the enemy. For example, if a map features uneven terrain, placing ranged units on hills can increase their combat strength against melee combat units. This is because they would rst have to climb the hill (this is in this example presumed to be slower than walking on even ground, but is of course domain specic). This level also includes tactics using formations or splitting up the army to attack dierent positions. [Michael Chung, 2004] Strategic planning Strategic planning includes all kinds of high-level decisions such as when to build an army, what units to build, when to upgrade the technology, where to expand bases and where and when to capture important strategic places such as resources. This paper primarily focuses on techniques for strategic and tactical planning, because the strength of an AI is mostly determined by these levels. A decent unit control can be achieved with simple reactive agents consisting of a set of rules and corresponding behaviours. [Michael Chung, 2004] High level strategy RTS AI problems are hard to solve. Unlike chess, a game that can be solved using a variation of A* searching, real time strategy poses dierent more complex problems for Intelligent Agents. [Buro, 2003] Problems in real-time strategy games that have been identied include: [Buro, 2003] Resource management Players need resources in order to build and upgrade units and buildings. A good resource management agent is important for any kind of real-time strategy AI. The gathering and balancing of resources can be done relatively well with a simple reex agent, since the AI has all the relevant information. It can simply assign more units to a certain resource when it runs low. [Danny C. Cheng, 2004] Decision making under uncertainty In an envrironment where only partial information is available and a certain strategy does not always produce the same result,it is much harder to reason. To be able to recognize important events the game probably needs some sort of pattern recognition. [Timo Kaukoranta, 2003] Even with no concrete information available, the AI should be able to form realistic hypothesizes and plans to execute. [Michael Buro, 2004]

Study Tour Pixel 2010 - University of Twente
Chapter 1. Real-Time Strategy High-level Planning Spatial and temporal reasoning The world in an RTS changes fast. The applicability of a certain strategy needs to be constantly reevaluated. Terrain analysis is of the utmost importance in RTS games. Having a good spatial representation of the game world can signicantly improve the AIs tactics and strategies. [Kenneth D. Forbus, 2002] The hardest contraint for RTS AI is time. The world changes so fast that when an AI is calculating its strategy, it already became invalid. Collaboration Clearly lacking in RTS AI is collaboration. Computer players never work together with other players to overcome a strong opponent, while humans regularly form alliances to defeat stronger players. A good method for communication/observation between dierent AIs and humans is necessary.[Michael Buro, 2004] This sounds easier than it is, good collaboration means the ability to recognize the allied strategy and helping them in the right way. Opponent modeling, learning Even harder than nding the allied strategy is the ability to discover the strategy used by the opponent. Human players can spot weaknesses in a strategy and exploit them in a few games, while current AI players have no way of accomplishing this yet. Learning from previous experience and formulating applicable counter strategies is a problem to be addressed. [Buro, 2003] Others Other challenging AI problems are present in Real-Time Strategy games, such as path nding for large groups of units, but these techniques are not very important for a strategy module, so they are not further discussed. Real time strategy games are a challenging area of AI research with a lot of problems to address. Finding solutions for these problems might oer interesting algorithms that may be applicable to the real world.

Commercial Techniques

Some commercial RTS games are very popular. Games like Age of Empires by Ensemble Studios or Starcraft by Blizzard have sold millions of copies all around the world. Even though all of these games have some sort of AI, multiplayer where two or more human players ght against each other, is far more popular. This can mainly be contributed to the lack of decent articial intelligence in these games. [Buro, 2003] Reasons why current commercial RTS AI fails to challenge human players are: Predictability and the lack of spatial reasoning Human players excel at nding weak spots in strategies and using the terrain to its limits and nding suitable strategies to exploit them. Commercial strategy AIs only spatial reasoning is in the path nding algorithms. [Kenneth D. Forbus, 2002] This weakness combined with their predictability makes it very easy to win against these kinds of opponents as a human. The predictability of commercial AIs comes from the fact that all AI is scripted. Even though these scripts can be really complex with lots of conditions and random events the AI is not able win the game in a fair way even against mediocre human players. Scripting A scripted AI is essentially a very complex set of rules that are predened by an expert in the domain. In RTS this means that all paths of attack, all buildorders and all tactics are static and predened. A buildorder is a sequence of actions that are considered good. For example when a game starts, rst build 5 resource gathering units and let them gather food (a resource needed for new units). A fully implemented script often has a way to tell the dierence between the begin, mid and endgame, where dierent parts of the script are active. At the start of a game buildorders are needed, while in mid-game for example the AI has to focus on making an army and expanding, while in the late game the focus lies on conquering and destroying the enemy. We could call such a complete script a Strategy. Commercial programmers include a set of these Pixel Articial Intelligence Indepth research 3

strategies in their games to give the human player some diversity. Even though these scripts have several disadvantages as discussed below, they are still the only technique used in commercial RTS games. This is because it still does give a static and good performance and is quite hard to beat even with recent Academic techniques. Hard to implement The implementation of a scripted RTS AI needs several experts in the game that think of viable strategies. Even after implementation it has to be tested and tweaked thoroughly. In combination with limited funds for development in commercial games and the fact that the rest of the game has to be near completion before the AI programmers can start implementing, causes that the available time to make and test commercial RTS AI is limited. [Michael Buro, 2004] Domain specic After implementing a scripted AI for a certain game, the implementation is not applicable for any other RTS games. This forces game developers to go through the implementation cycle again when designing the next game. If games are very similar it might be possible to reuse some of the old code, but generally a new game equals a new scripted AI [Istvn Szita, 2007]. Game developers use other means to entertain the player. Commercial games usually ship with a few dierent scripted strategies, so that it seems the AI does not always do the same thing. Another method is cheating, they give the AI more information than it should have. For example information about the players army even though it is not visible, or complete map information even though they did not discover it yet. Even with all this extra information most AIs are still not strong enough to compete against expert human players. Game developers can give the AI additional advantages in terms of higher resource acquisition rate, or stronger units. [Various, 2002] Having to rely on cheating to even pose a threat to expert human players is a clear indication how weak the RTS AI in modern day commercial games really is. The AI created with these cheats can be quite formidable for casual players, often the best AI is made weaker for novice human players.

Academic Research

In the past few years some researchers have tried to improve the performance of RTS games. Unfortunately the number of researchers is not yet quite high, as RTS has not really been brought to light as the new AI challenge as of yet. A very popular and related area is robotic football. It is related since it also features real time decision making and planning. Robotic football however is also focuses a lot on other Articial Intelligence challenges, such as visual recognition and robotic-movement functions. On purely tactical and strategic level robotic football is a lot less complex than even a simple RTS, as robotic football does not feature a economy, nor does it have interesting terrain features. Agents in robotic football also have almost complete information about the environment. The most important dierence is in the amount of objects. RTS games feature such a large number of objects and states that new planning methods have to be explored. [Buro, 2003] RTS games that have been used in research are Wargus and ORTS. Wargus is a Warcraft II clone based on a open source engine called Stratagus. It is a good candidate for AI research because its code can be easily modied and extended. [David W. Aha, 2005] ORTS is an RTS game engine specically designed for research. It has some interesting features, for instance the server executes the actions given by clients and sends only the available information for that client back. In that way there is no way to cheat, as the client has no way to know more information than is given by the server. Furthermore there is no default unit behaviour. So the player will have to give all the low level unit commands.[Michael Buro, 2004]

Dynamic Scripting

Commercial RTS games use scripted AIs. One of the biggest disadvantages of a scripted AI is that it is always using the same strategy and is therefore predictable. To address this problem Istvn Szita [2007] focused on Dynamic scripting to bring more diversity into the game. Dynamic scripting is essentially a reinforcement learning technique for scripts. [Istvn Szita, 2007] Dynamic
Chapter 1. Real-Time Strategy High-level Planning scripting automatically generates its strategy on the y by selecting a viable tactic in the tactic database. Each of these tactics is predened and domain specic. Every selectable tactic is given a certain weight. When a tactic is used against a player and it had a positive eect, its weight increases and vice versa. The sum of all the weights in the tactic database is constant, so a certain function is responsible for increasing and decreasing weights on tactics. [Marc J.V. Ponsen, 2005] Against static (using the same tactics every match) opponents dynamic scripting develops a strong counter-tactics that can beat most static opponents. Dynamic scripting can address problems on all three abstraction levels, as opposed to some previous research that for example applied reinforcement learning to low level unit behaviour. [Marc J.V. Ponsen, 2005] Advantages of dynamic scripting Whereas static scripts may contain weaknesses that can be found and exploited by human players, a dynamic script automatically reduces the eectiveness of these exploits, possibly making the game more interesting for human players. A dynamic script as proposed by Marc J.V. Ponsen [2005] has a small disadvantage. If it is trained against static opponents its strategy converges to a static and predictable strategy. Losing its strong point of being interesting. To further increase the diversity in tactics Marc J.V. Ponsen [2005] proposes a certain diversity function that rewards the AI for nding eective tactics that signicantly dier from the previous tactics. This way the adaptively of the strategy can even be increased as the AI considers more options, making them stronger. Typically, this technique leads to a more interesting game for human players. [Istvn Szita, 2007] Disadvantages of dynamic scripting Most research done in dynamic scripting uses static opponents to learn from. But it is much harder for an adapting script to become eective against human players since the opponent also adapts and nds weaknesses in the new strategies. Thus its eectiveness against other adaptive players can be doubted. [Istvn Szita, 2007] Furthermore, making dynamic scripts still requires experts and domain knowledge to ll the tactics database. Making a decent AI would still require a lot of implementation time, probably even more than static scripting.

Monte Carlo Planning

A dierent approach to the strategic planning problem is Monte Carlo planning. Monte Carlo planning essentially generates a set of simulations, for all possible actions. It then chooses the plan that corresponds with the best simulation for the player and acts upon it. Note that in an RTS game the number of possible actions can be very large, so large that it might be impossible for modern computers to look even a few steps into the future. That is why an abstraction for actions and states has to be found. Typically, a larger number of considered plans results in a better strategy. [Michael Chung, 2004] A plan in Monte Carlo planning consists of a sequence of repeating high level actions, such as attack. A large amount of these plans is generated, the outcome simulated and the best plan found. To nd this plan there needs to be some sort of evaluation function that rewards eective plans. The quality of this evaluation function is important for ecient good results, but a less good evaluation function can still perform good by increasing the search eorts. [Michael Chung, 2004] In essence the Monte Carlo planning algorithm focuses on adversarial planning. Anticipating what the enemy will do and constructing a viable counter strategy for it. Advantages of Monte Carlo planning An important advantage of Monte Carlo planning is that it is less dependent on domain experts to make predened tactics. Only the high-level actions and evaluation function are domain specic, and the majority of the evaluation function can be reused. [Michael Chung, 2004] Disadvantages of Monte Carlo planning Monte Carlo planning introduces a lot of calculating. In large complex games the amount of calculating could limit the depth of search the Monte Carlo algorithm can do. To keep the amount of considered plans and depth the same, even higher abstraction levels would be needed. [Michael Chung, 2004] Another problem with Monte Carlo planning is that it does not learn from
Pixel Articial Intelligence Indepth research
previous mistakes. If it is playing against a learning adversary, and he nds a winning strategy, it will not be able to improve itself.

Case Based Planning

Case Based Planning (CBP) is a planning implementation based on case based reasoning. Case Based Reasoning (CBR) is similar to the dynamic scripting method as it also looks at the past and learns from mistakes. CBR is based on states. It looks at past experience and calculates the best sub plan in the current state. Current case based planning AIs focus on tactical and strategic levels of operation with high level abstract state spaces. The CBP algorithm is given an initial goal: winning the game. This goal is than expanded to nd the sub goals that need to be completed rst. The Case Based planning algorithm given by David W. Aha [2005] simply selects the best sub plan given the current state and goal. If the execution of the plan was successful and eventually leads to victory, the choice of the sub plan is considered good, and the heuristic value of that certain sub plan in that state is increased. Case Based planning does not do any adversarial planning, nor does it try to model the opponent. CBP also uses a database of predened tactics. Initially, when learning what tactics should be used in dierent states, the algorithm picks a random one and remembers what outcome it has. Obviously, good tactics will get high evaluation values and vice-versa. [David W. Aha, 2005] Instead of pre dening strategies, Manish Mehta [2007] explains a way of learning from observing experts that play the game. An expert plays the game against a opponent (human, or AI) while the game makes a trace of all the actions done by the expert. When the game is nished the expert will indicate for each action what goal he tried to complete with it. The algorithm by Manish Mehta [2007] then computes viable tactics in certain states. The CBP algorithm can learn from its mistakes by doing certain bad tactics less often, but it has no way of xing the tactics itself. A further improvement on the Case Based planning algorithm is given by Manish Mehta [2009]. They tried to x exactly that problem. When plans fail, this information is remembered and when the system is not playing, it can calculate what the dierences are between successful plans and failures. When playing a game, the failure detection agent uses failure-patterns to discover causes of plan failures. When a explicit cause of plan failure is discovered, the corresponding tactic can be xed using a so-called plan-modication routine. [Manish Mehta, 2009] The main dierence between Case Based Planning and Dynamic scripting is the underlying idea. In Case Based Planning the strategy and tactics consist of sub plans and sub goals and it looks if the goals set are reached. In Dynamic scripting there is no such thing as goals. It simply selects the tactic that yielded the best results in the past. While the result of both algorithms may be the same, the underlying principle is very dierent. [David W. Aha, 2005] Advantages of Case Based Planning An advantage of case based planning is the implementation time, by allowing the experts to play the game and show the AI what good decisions are, simplifying the process of designing an AI. [Manish Mehta, 2007] CBP is able to form eective strategies against dynamic opponents and can provide some diversity as well. [David W. Aha, 2005] Disadvantages of Case Based Planning A disadvantage of CBP based AI is that it needs to be trained against opponents before it can become strong. The more complex and diverse the opponents are the longer it takes CBP to learn how to win. [David W. Aha, 2005]

Others methods

There are many more approaches and extensions to the methods mentioned here. An example is transfer learning using CBR and reinforcement learning. It focuses on the application of previous experiences and tactics in new unknown states or maps to reduce the training time needed for CBP algorithms to work eciently. [Manu Sharma, 2006] Totally dierent RTS AIs or parts of 6 Study Tour Pixel 2010 - University of Twente
Chapter 1. Real-Time Strategy High-level Planning RTS AIs could be designed using other algorithms. Like deductive, abductive or probabilistic (using Bayesian networks) plan recognition, but will not be further discussed in this paper since they are not yet applied to RTS games. [Danny C. Cheng, 2004]

Discussion

The availability of a open source RTS AI development platform, ORTS, will probably give the RTS researchers a way to test results against each other. It is kind of strange to nd that many researchers still using the game Wargus for research. Furthermore there really should be an international RTS AI tournament, like there is for robot soccer, to get some more attention to this domain. Applicability in other domains The American Military wants High-performance simulators for training military personnel. The level of command and control AI, in particular that of the RTS domain is lacking. [Michael Buro, 2004] To realistically train military personal in these areas, high performing AIs are needed. Furthermore, the problems addressed in section 1.3 can be applied to several other real-time domains, such as planning in real world scenarios. Current robotic research primarily focuses on recognition of the world rather than planning. Robotic football, a research domain that has some similarities to RTS, has almost complete information and its hardest challenges lie in knowing where you, your team and the ball are and calculating and communicating the strategy among the team. The strategy itself is a set of predictable attack paterns that could easily be countered by a human. Real-Time Strategy AI, could be applied here to nd more viable attack patterns. In general, the ability to plan in real-time is very important in a lot of domains. Which makes real-time strategy games a ideal test bed for complex real-world situations. [Buro, 2003] Would it be possible to combine several planning methods? To the best of my knowledge such a combination technique has not yet been used in the Real-Time Strategy domain, but it is certainly interesting what sort of results it would give. The biggest problems with such a technique is the amount of computing power it would need. Any of the academic techniques discussed in this paper already require vast amounts of computing power (except dynamic scripting). Combining them with for example a majority rule would require a separate calculation for each strategy and comparing them to nd the similarities. Which in turn would require a non-trivial strategy comparer that can identify the underlying tactic in each sequence of actions and nding the most preferred tactic. Although such a strategy comparer is already partially available in the Case-Based Planning technique. Combining dierent techniques on dierent command levels (as explained in section 1.3) has been done in research before though. The lowest level of unit control has a quite good performance even with simple reex-agent scripts, so the already available scripting has been used in some research to provide the most basic unit control. It is imaginable that for example Monte Carlo planning could be quite eective at the tactical control level, where the eect of ones actions can be calculated quite eciently since we reduced the state-space to the military control. While for example Case-Based Planning would take care of the high-level strategic plans. Such a strategy might take advantage of the strong points of both techniques while minimizing the weaknesses. Many more combinations of techniques are possible and it would be interesting to see the results of these combinations.If combined in the right way it hopefully creates an AI system that uses certain algorithms where they have the best results. There is a chance, however, that certain AI modules would have their own strategy that diers from the general strategy outlined by the system, which could result in conicting priorities and overall worse performance.

Conclusion

In this paper we have discussed the various real-time strategic planning AI techniques currently in use in commercial or research projects. The most prominent are the commercially much used scripting techniques, an academic method that dynamically chooses between scripts, a Monte Carlo Pixel Articial Intelligence Indepth research 7
planner and a Case Based Planning system. Pointing out which of these techniques is currently the best is quite hard. As there have been made some really advanced scripts in commercial games and the performance of some Case Based or Monte Carlo planners have yet to be proved in more complex games. The current academic technique that is the most advanced would be Case Based planning since it already can adapt, change and enhance available strategies and can be extended with a adversarial plan recognition system. [Manish Mehta, 2007] That said, there have not yet been any real performance comparisons between dynamic scripting and Case Based planning. [David W. Aha, 2005] Neither have there been any comparisons between Monte Carlo planning and either of the other two. Although Monte Carlo does not learn from its own mistakes, it might become a formidable opponent. Further research in this area will be needed before any of them come close to a human expert.

References

Buro, M. Real-Time Strategy Gaines: A New AI Research Challenge. 2003. Danny C. Cheng, R.T. Case-Based Plan recognition for Real-Time Strategy Games. 2004. David W. Aha, Matthew Molineaux, M.P. Learning to Win: Case-Based Plan Selection in a Real-Time Strategy Game. 2005. Istvn Szita, Marc Ponsen, P.S. Interesting Adaptive Game AI. 2007. Kenneth D. Forbus, James V. Mahoney, K.D. How qualitative spatial reasoning can improve strategy game AIs. 2002. M. Buro, T.F. RTS Games as Test-Bed for Real-Time Research. 2003. Manish Mehta, Santiago Ontanon, A.R. Case-Based Planning and Execution for Real-Time Strategy Games. 2007. Manish Mehta, Santiago Ontanon, A.R. Using Meta-Reasoning to Improve the Performance of Case-Based Planning. 2009. Manu Sharma, Michael Holmes, J.S.A.I.C.I.A.R. Transfer Learning in Real-Time Strategy Games Using Hybrid CBR/RL. 2006. Marc J.V. Ponsen, Hctor Muoz-Avila, P.S.D.W.A. Automatically Acquiring Domain Knowledge for adaptive Game AI Using Evolutionary Learning. 2005. Michael Buro, T.M.F. RTS Games and Real-Time AI Research. 2004. Michael Chung, Michael Buro, J.S. Monte Carlo Planning in RTS Games. 2004. Timo Kaukoranta, Jouni Smed, H.H. Role of Pattern Recognition in Computer Games. 2003. Various. AI Game Programming Wisdom. 2002.