(print Project)

Cs472 Project: Planning and Dialogue Generation for Gaming

This is a Project page, written Fri Jan 11 13:01:07 PST 2008.

Notes

Project size: three to four persons per group. And project1 for cs472 is the same as cs572.)

Parts

Project 1: An introduction to LISP programming
Project 2: Some planning
Project 3: Some dialogue generation controlled by the planner

Background

Games needs graphics and models:

Graphics to make it look exciting and give feedback;
models to control what the computers offers you.

Other subjects deal with the graphics. Here, we deal with the model.

Our game concerns the ancient and stressful act of getting married. Not everyone gets married and some of us never should. It just does not work for some people:

Marriage is a fine institution but who wants to live in an institution?
-- Gaucho Marx
Sometimes I wonder if men and women really suit each other. Perhaps they should live next door and just visit now and then.
-- Katerine Hepburn

Those of us who do get married, need all the help they can get.

The Wedding Planner

In this game, our AI agent will be like an on-line wedding planner, advising a budget conscious couple about, well, everything.

Note that the goal of producing a "great wedding" is a non-linear problem: produce the best wedding possible using the least cost. Not all is possible. Trade-offs will have to be made. One problem is all this is the dependencies between decisions. For example,

choice of guests may effect choice of menu
which effects choice of wine
which effects total cost
which effects how many guests are invited
which effects who gets invited
which effects menus and around we go again.

Your tool should report N different wedding plans and plot the cost/benefit of each.

Also note that the wedding planner and the ?happy couple are not always sharing the same goals. The wedding planner may have left over stock of some material they want to sell to the couple- but the couple may not be convinced that they need it.

Note that your tool has to be an inference engine and a knowledge base. Part of the art of AI programming is showing that the same inference engine can run over multiple knowledge bases. So pick any two of the following weddings and try to model some of the rituals.

References

Chinese weddings: see The Wedding banquet (Chinese wedding that gets a little... strange).
Indian weddings: see Monsoon wedding (huge, contemporary wedding).
British weddings: see Four Weddings and a funeral (pick any one of the weddings and ignore the bit about the funeral)
Greek weddings: see My big fat Greek wedding (before the wedding, don't forget all the courting).

Note: you won't be able to model all of any of these. To make the task manageable, pick some part of the whole problem; e.g. try to model one of banquets in the movie.

Cs572 class project: Model-based software engineering

This is a Project page, written Fri Jan 11 13:31:40 PST 2008.

Notes

Project size: two people per group. And project1 for cs572 is the same as cs472.).

Parts

Project 1: An introduction to LISP programming
Project 2b: Certifying the models
Project 3b: Implementing the search engines
Project 4b: Experimental studies

Background

The STAR tool, discussed in class, is a decision support aid for model-based software engineering.

STAR has a search bias. It only explores options using a very dumb search device called simulated annealing. Your task is to check how different biases change the results of that tool

Task

measured in terms of

Runtime
Size of solution
Median score of the solution
Variance in the solution score
The stability of the solution over multiple runs.

For all the above except stability, less is more (i.e. the lower the numbers, the better).

You will explore the above using simulated annealing and any three of the following:

Keys
Depth-first iterative deepening
Isamp
Beam search
LDS

(And, for extra credit, explore more than 3 including items not on this list.)

References

The Business Case for Software Engineering
The Keys paper
Simulated annealing
The LDS paper (has some notes at the front about Isamp).

Reports

Reports: You will hand in a PDF (ten pages make) document in format of the ACM Sig proceedings: template.

Project 1: LISP, Grammars

This is a Project page, written Thu Nov 22 22:45:32 EST 2007.

(Note: do not attempt this project till you have completed Lab1.)

Part A: A whole lotta LISP

In Lab1, you were shown how to:

create a "make.lisp" file that controls a set of files
write test cases using "(egs :topic)" and run test cases using (demo :topic)

Your task in this project is to apply that knowledge to show you understand Norvig's LISP code in chapters 1,2,3. For each chapter:

Create a separate directory "proj1/chX"
Create a different "make.lisp" file in each directory that only loads files in that directory
Enter in code from Norvig chapter X
Create one "(egs :X)" test suite for each second-level heading that you don't skip over. With each eg, write 2 lines explaining what is going on here and include an expected output.
Create a "demos.lisp" file that contains "(egs :all)" that calls all the other "egs" (if you don't understand this bit, go back and look at the lab1 "demos.lisp" file).

Only do code from the following sections:

1.1, 1.2, 1.4, 1.5, 1.6, 1.7, 2.2, 2.3, 2.5, 3.1, 3.3, 3.5, 3,6, 3.8, 3.10, 3.12, 3.14, 3.15, 3.19

Part B: A Little Bit of Grammar

Modify the grammar on page 30 of Norvig to describe the sequence of cs/ee subjects required to get a cs/ee/ce degree at WVU.

Create a new directory "proj1/degree"
Create a file "proj1/degree/make.lisp" file.
Install just enough of the chapter 2 code into this new directory so the grammar can run.
Create example code in "proj1/degree/demos.lisp#(egs :all)" which, if executed, shows three different student plan of studies.
Comment your code.

What to hand in

Comment your code

Zip up the entire "proj1" directory to "proj1.zip"

Submit the zip to Ecampus.

How this code will be tested

With your whole group in attendance...
I will unzip your zip
In each directory, I will run "(load "make.lisp") (egs :all)" (in SLIME, on a CSEE LINUX box).
For each member of the group in turn, I will point to ten random parts of the code and ask them to explain it to me (no hints from other members of the team). I will expect you to know your code.

Study hint

Any code in sections 1.1, 1.2, 1.4, 1.5, 1.6, 1.7, 2.2, 2.3, 2.5, 3.1, 3.3, 3.5, 3,6, 3.8, 3.10, 3.12, 3.14, 3.15, 3.19 may be asked in the mid-session quiz.

Exams, simpler assignments (building a conversation agent for a game);
No exams, elaborate assignments (building an expert system for software project management);

All CS472 students with a GPA below 3.5 must use scheme A.

All advanced AI students (CS572) students must use scheme B.

All other students get to choose, provided the lecturer gives permission.

week	A (with exams)	B (with no exams)	Marks
4	Project1	Project1	10
7	Mid-term	Project2b	20
10	Project2a	Project3b	20
13	Project3a		20
14		Project4b (presentation)	20
17	Final exam	Project 4b (written submission)	30

Final marks:

Late marks

Assignments can be submitted up to three days after the due date, at a late penalty of 2.5 marks per day.

After 3 days late, assignments will get a grade of zero.

(print Project)

Cs472 Project: Planning and Dialogue Generation for Gaming

Notes

Parts

Background

The Wedding Planner

References

Cs572 class project: Model-based software engineering

Notes

Parts

Background

Task

References

Reports

Project 1: LISP, Grammars

Part A: A whole lotta LISP

Part B: A Little Bit of Grammar

What to hand in

How this code will be tested

Study hint

Project 2a

Project 2b

Project 3a

Project 3b

Project 4b

Grading

Late marks

cs472 / cs572

(print *Project*)

Notes

Parts

Background

The Wedding Planner

References

Notes

Parts

Background

Task

References

Reports

Part A: A whole lotta LISP

Part B: A Little Bit of Grammar

What to hand in

How this code will be tested

Study hint

Late marks

(print Project)