
Explore how intelligent agents perceive environments with sensors and act through actuators. Learn how rational agents map percept sequence to actions to maximize performance, and how autonomy emerges through learning.
Introduction to Game, Common Games in Artificial Intelligence Research, Key Concepts - Tic- Tac- Toe Game, Game Tree Search Tree
Apply minimax to a tic tac toe position for the max player to choose next move, analyze five options, and identify the only favorable move that blocks O’s immediate win.
Study the pursuit and evasion game on a city graph, where P pursues E along edges in turns, and payoffs equal the time to meet, via a game tree.
Examine a four-room, two-player game with A at 1 and B at 4; build the game tree, mark terminals, and illustrate loops with a question mark; leaves +1 or -1.
Optimal Decisions in Games - Two-person Zero-Sum Game - Min-Max Algorithm
Apply the min-max game algorithm on a rooted tree with alternating max and min layers; evaluate bottom-up to obtain the root value, which is nine.
Multiplayer Games, Utility Function, Win amount for each player, Alliances and partnerships among players
Learn how alpha-beta pruning speeds up game tree search by pruning branches in minimax, updating alpha and beta to avoid exploring inferior moves and reduce computing time.
Alpha-Beta Pruning and Move Ordering
Apply the min max algorithm with alpha-beta pruning to determine the root value, showing which branches are pruned.
Imperfect Real Time Decisions
Cutting off search, Forward pruning
Stochastic Games - When the outcome does not depend upon anything and depends only on chances, it is called a stochastic experiment. A game with a mix of a stochastic experiment and some moves is called a Stochastic Game. Backgammon Game - Schematic Game Tree
Explore alpha-beta pruning on a game tree, evaluating leaves left-to-right and updating alpha and beta to determine the max's best move; demonstrate pruning decisions and when to stop evaluating leaves.
Explore partially observable games and the fog of war, with examples like Battleship, Stratego, and Kriegspiel. Learn how percept sequences build belief states for strategic decision making.
State of the art Game programs - Deep Blue, HYDRA, RYBKA, Checkers / Draughts, Othello / Reversi, Backgammon, The GO Game, MOGO, Combinatorial Game Theory, Card Game BRIDGE, Scrable, QUACKLE
Alternate approaches for developing Artificial Intelligence based Game programs
The objective of this course is to teach the concepts, principles, and practices used to develop artificial intelligence-based game programs.
Games are played by people to prove the intelligence and skill of the players. As Artificial Intelligence spans across domains to build machines as capable as humans, both physically and cognitively, Artificial Intelligence-based Games programs is one of the emerging areas in technology. In 1997, the first Game playing machine, IBM Deep Blue defeated the Russian Chess Grandmaster, Garry Kasparov it was looked at as a historic beginning of the competition between machines and man. In this course, we begin with a brief introduction to the structure of an Intelligent Agent. We begin with a simple two-person Min – Max zero-sum game in which the win amount of one player is the loss of the other. The min-max algorithm aims at exploring all possibilities estimating the worst-case guaranteed amount and choosing the move. We also extend to multiplayer games where the utility function is a vector of dimension k (number of players). Alpha Beta Pruning increases the speed in decision making, by pruning some unwanted branches of the search tree. As the number of nodes to be explored is exponentially many, we try to develop some meaningful heuristic approach to develop an algorithm to cut off a few branches and get some imperfect real-time solutions. Skill-based games are played purely with the skills of players. Stochastic games combine some stochastic experiments to decide who will do the next move and how. We also deal with stochastic games and the algorithms to choose the next move. Some games are fully visible by both the players while in some games, a player's board is hidden from the other. Partially observable games use statistical methods to arrive at meaningful solutions. We also discuss some state-of-the-art game programs and modern approaches.