ai lab-1

download ai lab-1

of 2

Transcript of ai lab-1

  • 8/11/2019 ai lab-1

    1/2

    AI Lab

    Assignment 1: GOAL

    General Instructions:You can use any programming language that you like, provided it is available at thesystems you are using or can manage one. The submissions need to be finally made to the TAs.

    Please discuss with the TA regarding the form of submission (only soft copies allowed, nothing

    in hard copy). A dedicated team of TAs will check a large number of random pairs of codes. Anyformat of cheating and copying would go to the instructors for action. Minor copying would be

    laughed at and ignored, serious ones would be dealt with the same degree as copying in any

    written examination as per institutional policies (barbarically, if it does not exist). To avoid

    problems please:i) Make sure your roll number and name appears in all source code text files that you make (1st

    line of all source files) along with the assignment/question number.

    ii) Any material copied from a friend or online (if permitted or necessary to proceed) is clearlymarked (// proudly coped from XXXX, he's awesome)

    iii) Any non-standard external library or tool used is clearly marked in the source files at the start

    of the main file.

    The assignment is inspired by the FIFA World Cup, although having watched the matches or

    knowing the game to a specialized degree is not a requirement of the assignment. We willhowever take an overly simplistic system for this assignment. Consider a single player, a ball and

    a goal post. Consider that the field is made up of 10x10 blocks. Each block can either contain the

    player, the ball or both. The ball can only move along with the player, it would otherwise lie as

    static. Identify the Performance Measure, Environment, Actuators and Sensors for the systemand make a simulation architecture for this system, based on which answer the following

    questions.

    In all the questions make sure that the agent and the environment are kept separate, and each

    entity as separate as possible. Please submit at least 2 questions inside the lab. The solutions can

    be taken out, re-worked, extended, discussed and asked from the TAs/instructors. Extendedsolutions can be submitted separately, however the code part should be original or properly

    credited. Normally it should be OK to share architecture (with credit) but not the agents. Please

    maintain a separate project folder/source file for each question. Inputs can be hard coded in the

    source files or dynamically supplied at execution time. Please supply a text/word file for allquestions with the inputs and the final output (the path traced by the player and the performance

    measure). Repeat the same for 2 test cases per question. Only submit a question if the solution is

    fully correct and tested. There is no part marking.

    1. Imagine that the player can see the location of the ball and the goal post at all times, and his

    own location. The player and the ball occupy one block each, while the goal post is the

    region from (10,4) to (10,7). Move the player from his current position to the ball and thentogether with the ball to the goal post. A player can move in any of the 4 neighboring blocks

  • 8/11/2019 ai lab-1

    2/2

    using the action MOVE; and if the player is in possession with the ball, he can move along

    with the ball using the action KICK. For testing use different random locations of the player

    and the ball.2. Imagine that the player can only see the ball if that lies in a Manhattan distance of 3 blocks.

    Now modify the agent used in question 1 and build a state based agent. Note that there may

    be numerous ways to achieve this, any would suffice.3. Imagine that the player can only see the ball if that lies 'ahead' where from the direction thatthe player is pointing to. The player can point in any one of the 4 directions

    parallel/perpendicular to the field. Now modify the agent used in question 1 and build a state

    based agent. For this question the player may additionally turn at angles of 90 degrees andsense his current orientation.

    4. You realize that the only motion to the player and the ball is made by the agent itself and

    hence continuously sensing the positions of the two are not needed, rather the information

    can be maintained from the prior state of the player and ball and the actions applied to them.However now there is a goalkeeper who can exist in any one of the positions of the goalpost

    and cannot leave the post. The game ends if goalkeeper catches the ball. The goalkeeper, the

    player and the ball can move one block at a time. If a player/ball move to a location in thegoal post at the same time as the goal keeper, the priority is given to the player/ball. Make a

    random agent for the goal keeper which randomly moves the goalkeeper.

    5. Add 2 opponent players, who cannot move. Their positions at any time can be sensed. Make

    a goal based agent which optimally moves the player to the goal. Solutions based onassumptions would be acceptable as long as the assumptions are stated.

    6. Repeat question 5 when the players can move. However if any two players simultaneously

    aim to reach the same position, they collide and the game is stopped. The game resumes withthe players and the ball in the original position (before colliding) and the ball in possession of

    the player initially with the ball. Make random agents for each of the players. Solutions based

    on assumptions would be acceptable as long as the assumptions are stated.