1 / 20

CSE 380 – Computer Game Programming Artificial Intelligence

CSE 380 – Computer Game Programming Artificial Intelligence. Murloc, Blizzard Entertainment. Artificial Intelligence. Can a computer “think”? Can it process info in a fashion similar to humans? How do we make creatures “appear” to think? We only want creatures intelligent enough

guido
Download Presentation

CSE 380 – Computer Game Programming Artificial Intelligence

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. CSE 380 – Computer Game ProgrammingArtificial Intelligence Murloc, Blizzard Entertainment

  2. Artificial Intelligence • Can a computer “think”? • Can it process info in a fashion similar to humans? • How do we make creatures “appear” to think? • We only want creatures intelligent enough • The ultimate AI game – massively multiplayer • The ultimate stupid programmer: • Press Your Luck – beaten by contestant Michael Larson • http://gscentral.net/larsen.htm

  3. Remember Our Infrastructure GameStateManager 1 1 * SpriteManager AnimatedSpriteType 1 1 1 * Bot 1 1 AnimatedSprite CollidableObject PhysicalProperties 1 * BoundingVolume

  4. Bot.h class Bot : public AnimatedSprite { public: Bot(){} ~Bot(){} // TO BE DEFINED BY BOT AI CLASSES virtual void think(Game *game)=0; virtual Bot* clone()=0; };

  5. BotRecycler? class BotRecycler { private: map<wstring, list<Bot *> *> recyclableBots; map<wstring, Bot *> registeredBotTypes; public: BotRecycler(); ~BotRecycler(); void addMoreBots(list<Bot *> *botsNeededList, wstring botType, unsigned int numToAdd); void initRecyclableBots(wstring botType, unsigned int maxBotsNeeded); bool isRegisteredBotType(wstring botType); void recycleBot( wstring botType, Bot* botToRecycle); void registerBotType(wstring botType, Bot *sampleBot); Bot* retrieveBot(wstring botType); };

  6. Deterministic AI Algorithms • Behaviors that are predetermined or preprogrammed • Objects may follow a straight course • Asteroids • Random Motion • Good for modeling things like insects • Periodically (not every loop iteration), pick random direction & velocity for object(s) • Weight probability in a particular direction for the wind • Tracking Algorithms • Takes into consideration the position of object being tracked • Change path directly toward object, OR • More realistically, gradually change path toward object

  7. Direct Tracking Algorithm • Direct Tracking – like the terminator • Example: if (player.x > monster.x) monster.vX = MAX_V; if (player.x < monster.x) monster.vX = -MAX_V; if (player.y > monster.y) monster.vY = MAX_V; if (player.y < monster.y) monster.vY = -MAX_V; • Not very realistic because tracking is so precise

  8. Trajectory Vector Tracking • Gradually change the path toward the target • Compute the vector from the tracker to the target • TV = (target.x-tracker.x, target.y-tracker.y) = (tvx,tvy) • Normalize: (tvx,tvy)/Vector_Length(tvx,tvy) • Now max length is 1.0 (remember physics?) • Adjust the current velocity vector of the tracker by adding TV* scaled by a rate of your choice • tracker.vx += rate*tvx; • tracker.vy += rate*tvy; • If velocity has overflowed a maximum rate, slow it down

  9. Patterns & Basic Control Scripting • Creatures perform a series of actions in a particular order • Motion Control Patterns • Used in Space Invaders, Galaxian, Phoenix, etc … • Create a pattern instruction set, like: #define GO_FORWARD 1 #define GO_BACKWARD 2 #define TURN_RIGHT_90 3 #define TURN_LEFT_90 4 #define SELECT_RANDOM_DIRECTION 5 #define STOP 6 • Use another operand to keep track of how long to do each action • Use an array to store the order for a pattern

  10. Pattern Example • Globals: int num_instructions = 6; int square_stop_spin[1, 30, 3, 1, 1, 30, 3, 1, 1, 30, 3, 1, 1, 30, 6, 60, 4, 8]; • During Game Play: int instruction_ptr = 0; int cycles = spare_stop_spin[instruction_ptr+1]; switch(square_stop_spin[instruction_ptr]) { case GO_FORWARD: //move creature forward … } instruction_ptr += 2 if (instruction_ptr > num_instructions*2) // sequence over, perhaps do again

  11. Improving on Patterns • Make sure you provide feedback control to your pattern engine so your monster doesn’t do anything illegal • No collisions, going off screen, etc … • Patterns with Conditional Logic Processing • You can choose from a number of patterns randomly • To make the AI smarter, choose a pattern based on the current game conditions • Test for distance to player, when the player gets closer, change the monster pattern

  12. Behavioral State Systems • To create a Finite State Machine (FSM), you need: • A reasonable number of states, each which represents a different goal or motive • Example Monster states: • Move forward, move backward, turn, stop, fire weapon, chase player, evade player • A single state may itself have multiple sub-states • Lots of input to the FSM, such as the state of the environment & the other objects within the environment • Continually update the monster so they are behaving on the latest state • You can model aggression, curiosity, stupidity, etc …

  13. Elementary State Machines • Create a master brain that deals with high level states, which then get translated down to lower states • Ex: High-level creature behavior states may be: • Attack, retreat, move randomly, stop, look for something, select a pattern & follow it • For last state, a pattern processor selects a pattern of behavior for the creature based on environment

  14. ESM Example switch(creature_state) { case STATE_ATTACK: // attack logic case STATE_RETREAT: // retreat logic … case STATE_PATTERN: Process_Pattern(); } creature_state = rand()%6; • We could give more personality to monster by creating a distribution of behavior. Ex: State Goomba p(x) Koopa p(x) ATTACK 50% 15% RETREAT 20% 40% STOP 5% 45% RANDOM 25% 5%

  15. Modeling Memory & Learning • So far our creatures think on-the-fly • Creatures can learn from past experiences (if you want them to) • Don’t fall for the same player moves • Creatures searching for something may remember where they’ve been • Many game programmers like to use bit strings or vectors to memorize data

  16. if x1 op y1 false true if x5 op y5 false true if x2 op y2 false true if x3 op y3 false true if x4 op y4 false true Planning & Decision Trees • Planning is high-level AI • Using a high-level set of actions that are carried out to arrive at a goal • There may be conditionals that must be satisfied before any particular action can be carried out • Production rules can be used in decision tree:

  17. Decision Tree Example Conditions • If the player is close and damage is low • Then fire at player • If the player is far and fuel is high • Then search for player • If damage is high and player is close • Then evade player • If damage is high and ammo is 0 and player is close • Then self destruct • You may want to define structures that specifically represent decisions to be made

  18. Pathfinding • In our earlier example, there were no obstacles for vectoring • If there are obstacles, things get complicated • Resolutions • Trial and error – if a collision occurs, back-up the monster, turn it 45 degrees and try again • Contour Tracing – trace the contour of the object that’s blocking the path • Collision Avoidance Tracks – create virtual tracks around objects that trace out a fairly intelligent path • The path can be computed algorithmically using a shortest-path algorithm • Waypoint Pathfinding – for complicated world with many obstacles, setup a network of paths • Creatures only need to find their way to nodes, which are then connected via paths • All sorts of CSE 373 & AMS 301 algorithms could be used for finding shortest paths

  19. Advanced AI Techniques • Advanced AI Scripting • Examples: QUAKE C, UNREAL Script • Allow you to program game code with a high-level English-like language that is processed by the engine • You may define your own scripting language • A model of the human brain • Brain has 10-100 billion brain cells, on which end are you? • Genetic Algorithms • A method of computing that relies on biological models to evolve solutions • Fuzzy Logic • Making deductions about fuzzy set theory • Analyzing sets of data such that the elements of the sets have partial inclusion (objects are partially in a set)

  20. Basic Guidelines for AI • Objects with simple behaviors (rocks, missiles, etc …) • Use Deterministic AI • Objects that are supposed to be smart, but are part of the environment rather than the main action (birds that fly around, a space ship passing by, etc …) • Use deterministic AI coupled with patterns & randomness • For important game characters that interact with the player • Use FSM coupled with other supporting techniques like personality distributions • Main characters should be very smart • Should be state-driven • Use conditional logic, probability, & memory for controlling state transitions

More Related