Windy Grid World

This assignment is to use Reinforcement Learning to solve the following "Windy Grid World" problem illustrated in the above picture. Each cell in the image is a state. There are four actions: move up, down, left, and right. This is a deterministic domain -- each action deterministically moves the agent one cell in the direction indicated. If the agent is on the boundary of the world and executes an action that would move it "off" of the world, it remains on the grid in the same cell from which it executed the action.
Notice that there are arrows drawn in some states in the diagram. These are the "windy" states. In these states, the agent experiences an extra "push" upward. For example, if the agent is in a windy state and executes an action to the left or right, the result of the action is to move left or right (respectively) but also to move one cell upward. As a result, the agent moves diagonally upward to the left or right.
This is an episodic task where each episode lasts no more than 30 time steps. At the beginning of each episode, the agent is placed in the "Start" state. Reward in this domain is zero everywhere except when the agent is in the goal state (labeled "goal" in the diagram). The agent receives a reward of positive ten when it executes any action {\it from} the goal state. The episode ends after 30 time steps or when the agent takes any action after having landed in the goal state.
You should solve the problem using Q-learning. Use e-greedy exploration with epsilon=0.1 (the agent takes a random action 10 percent of the time in order to explore.) Use a learning rate of 0.1 and a discount rate of 0.9.
The programming should be done in MATLAB. Students may get access to MATLAB here. Alternatively, students may code in Python (using Numpy). If the student would rather code in a different language, please see Dr Platt or the TA.
Students should submit their homework via email to the TA (suchismi@buffalo.edu) in the form of a ZIP file that includes the following:
1. A PDF of a plot of gridworld that illustrates the policy and a path found by Q-learning after it has approximately converged. The policy plot should identify the action taken by the policy in each state. The path should begin in the start state and follow the policy to the goal state.
2. A PDF of a plot of reward per episode. It should look like the diagram in Figure 6.13 in SB.
3. A text file showing output from a sample run of your code.
4. A directory containing all source code for your project.
5. A short readme file enumerating the important files in your submission.
Updates
You can initialize the Q function randomly or you can initialize it to a uniform value of 10. That is, you can initialize Q such that each value in the table is equal to 10.
There have been questions about how to know when the algorithm has converged. The algorithm has converged when the value function has stopped changing signficantly and the policy has stopped changing completely. Since we are using q-learning, the algorithm should converge to a single optimal policy.
Please also submit a short readme file with your homework that enumerates the important files in your submission.

Posted Date: 2/9/2013 2:30:26 PM | Location : United States

Ask an Expert

Related Discussions:- mathlab , Assignment Help, Ask Question on mathlab , Get Answer, Expert's Help, mathlab Discussions

Write discussion on mathlab
Your posts are moderated

Write your message here..

Related Questions

Texture - elements of composition, Texture Texture is how artists show... Texture Texture is how artists show what an object would feel like to touch. This can have a dramatic effect on how an image is perceived. Imagine for instance an image that h

What are the events used for page headers and footers, What are the events ... What are the events used for page headers and footers? The events TOP-OF-PAGE and END-OF-PAGE are used for pager headers and footers.

Two ways of producing a list within a transaction, What are the two ways of... What are the two ways of producing a list within a transaction? By submitting a separate report. By using leave to list-processing.

Mips assembly language equivalents , MIPS' native assembly code only has tw... MIPS' native assembly code only has two branch instructions, beq and bne, and only one comparison instruction, slt. Using just these three instructions (along with the ori instruct

C++ programming, how does a ohms law c++ programming works how does a ohms law c++ programming works

What are the general security issues, What are the General Security Issues ... What are the General Security Issues Many issues exist when linking a computer system to the Internet or indeed to an external link via a network set up. There are numerous way

Period and space tradeoffs - artificial intelligence, Period and Space Trad... Period and Space Tradeoffs - artificial intelligence: In practice, you are going to stop your agent for long term using in searching goal at some stage if it has not found a s

Micro programmed control and hardwired control, Explain micro programmed co... Explain micro programmed control. Ans: A micro programmed control unit is made around storage till is called a control store where all the control signals are stored in a prog

Explain about fundamental nature of a computer, Q. Explain about fundamenta... Q. Explain about fundamental nature of a Computer? The fundamental nature of a Computer is as an information transformer. So a computer should be able to take input and process

Explain about hybrid model, Q. Explain about Hybrid model? Hybrid model... Q. Explain about Hybrid model? Hybrid models are normally tailor-made models suiting to particular applications. In fact these fall in category of mixed models. These type of a