Project 0 (CSCI 360)

Project 0 (CSCI 360
3 Points

1 Introduction
Moving an agent from a starting state or location to a goal state or location is
one of the most fundamental tasks in Artificial Intelligence. In this project,
you will simulate a mobile robot performing this task on a discrete 2D grid
using a simple text-based simulator. Below is a screenshot of the text-based
simulator you will make use of in this course. Each dot represents a location
in the environment that is currently not occupied. The 0 represents the
location of the robot in the environment, while the $ represents the target
location. The goal is to use the actions and sensors of the robot to move it
from its current location to the goal location.
Figure 1: A screenshot of the text-based simulator. The environment is laid
out in a discrete grid. Each period represents an unoccupied location in the
environment. The robots location is represented by the 0, while the target
location is represented by $.
1
2 Simulator
2.1 Wheelbot
In this course, you will be working with a robot called Wheelbot, an abstract mobile robot encoded in the Robot class. At each simulation step,
Wheelbot can read its sensors and then take one action, moving it one space
in the given direction. For the purposes of this assignment, Wheelbot has
the following actions and sensors:
2.1.1 Wheelbot Actions
1. MOVE UP: Moves Wheelbot one grid space up (North)
2. MOVE DOWN: Moves Wheelbot one grid space down (South)
3. MOVE LEFT: Moves Wheelbot one grid space left (West)
4. MOVE RIGHT: Moves Wheelbot one grid space right (East)
5. STOP: Stops Wheelbot for one time step in the current grid cell
2.1.2 Wheelbot Sensors
1. Wheelbot Position Sensor: Provides the (x, y) coordinate of Wheelbot on the grid. To use, call this→simulator.getRobot()→getPosition().
2. Target Position Sensor: Provides the (x, y) coordinate of the target
on the grid (the $). To use, call this→simulator.getTarget().
3. Target Distance Sensor: Provides the Euclidean distance between
the target and the robot’s current location on the grid. To use, call
this→simulator.getTargetDistance().
3 Programming Portion
This project will require you to modify the function Project::getOptimalAction()
in the file Project.cpp in the source code provided for this project. You are
not to modify anything else in this file or any other file that is part
of the simulator. Feel free to look at Robot.h, which defines Wheelbot,
Simulator.h, which defines the environment, and Vector2D.h, which provides
2D vectors and points. We will test your project by copying your function
code into our simulation environment in the designated area and running it.
2
Any changes you make outside of the designated function will not
be saved or counted as part of your submission. Your code must be
standard C++ code that compiles using the provided CMakeLists.txt file
on Linux or MacOSX. For this project, you should not need any headers,
libraries, etc. outside of what is already provided in the Project.cpp file.
The requirements of your Project::getOptimalAction() function are as
follows:
1. Use the above sensors and actions to navigate Wheelbot from its random starting position to the target position.
2. Let a be the number of horizontal steps between the random starting
position and the target position in terms of manhattan distance. Let b
be the number of vertical steps between the random starting position
and the target position in terms of manhattan distance. Your robot
must take, at most, a + b steps for full credit.
4 Theory Questions
1. The parameter transitionProb at the beginning of the the function
main in the file main.cpp controls the determinism with which Wheelbot takes actions. If we set transitionProb ← 0.9, for example, with
probability 0.1 Wheelbot will perform an incorrect motion action (e.g.,
moving left instead of up). Run your solution code a number of times
with different values of this parameter (between 0 and 1) and explain
Wheelbot’s behavior. Does it always reach the goal in a + b steps (at
most). If not, why not? Explain how we might hope for our Wheelbot
to overcome this problem. Should Wheelbot ever hope to navigate
perfectly (in at most a + b steps) if transitionProb < 1.0?
2. How could Wheelbot be made to handle the presence of obstacles on
the grid (with known locations)? What if Wheelbot did not know
the positions of the obstacles in the grid but had a local sensor that
reported the existence of such obstacles when Wheelbot was one grid
space away from them. How could Wheelbot be made to handle such
”hidden” obstacles?
5 Submission
For this project, all you need to submit is your modified Project.cpp file.
Please submit your modified file by the due date listed at the top.
3