|
Post by Hannes Vilhjalmsson on Feb 17, 2011 8:36:13 GMT -5
Dear students,
Use this thread to propose final projects and advertise for group members. So, for example, if you have come up with an idea on your own, tell everyone about it here, both to get feedback about the idea and to see if someone else becomes interested in joining forces with you. You can be up to three people together with a final project.
Some ideas include:
* Expert system for a domain * Game player * Puzzle solver * Bot for an action game * Robot navigation (virtual or possibly real) * Chat bot with reasoning * Planning / scheduling system * Pattern recognition (e.g. for vision)
You do not have to build everything from scratch (there is not enough time for that), so you can typically build on top of an existing engine, for example, use an existing expert system engine but feed it with some domain specific knowledge and create a user interface.
Cheers,
- Hannes and Angelo
|
|
|
Post by kristjan on Feb 21, 2011 7:58:12 GMT -5
Im thinking about making a puzzle solver that solves sudoku. Gonna try making it in c#. People are welcome to join.
|
|
|
Post by Hannes Vilhjalmsson on Feb 21, 2011 10:44:01 GMT -5
This is a good idea Kristján. There are several different approaches to doing this, so you may want to do a bit of background research to find out what the general strategies are.
You are all free to build on existing work (and in fact encouraged to build on existing knowledge), but remember to include references to that work both in your code and report.
|
|
|
Post by niccolo on Feb 21, 2011 20:58:50 GMT -5
Hi, among other subjects, I am following Advanced Software Engineering where we study modeling and model checking reactive and concurrent systems, with a formal basis as well as a practical point of view. We also have to make a final project and I know that Prof. Marjan Sirjani has got some experimental tools to test.
if possible, I'd like to merge the two projects. I thought about the possibility to model check some AI. Looking at Internet, some time time ago, I found this paper: Verification of FSM using Attributes Definition of NPCs (digilib.unsri.ac.id/download/200607A2515082009.pdf). The authors of the paper designed and implemented NPC fish models and verified their property with Computational Time Logic. Then, I am looking at this stuff to understand if it fixes with my interests.
If comes out that that is good to make a similar study, I already have a 3D underwater virtual environment written in python (with Panda3D libraries) to test the results.
Another things which I thought they can need some AI and model verification could be unmanned spacecraft or manned vehicles with not-humans drivers (like the new subways..). That sounds interesting too!
Anyway, I still just looking at these things, so I am open to other ideas. Anybody interested in join me for the AI part of the projects is welcome!
K. Niccolò
|
|
|
Post by Elín Carstens on Feb 23, 2011 11:48:29 GMT -5
The main goal of our project is to make an agent, the ZenMaster, that can tell short zen-like stories. From 4 to 8 sentences aprox. We are going to use C# and a smile. If anyone wants to join us we have room for one more. Elín and Guðrún
|
|
|
Post by carmine on Feb 26, 2011 11:00:23 GMT -5
I would like to implement a program able to solve any kind of sudoku-puzzle like Kristján but I would like to use Java because I feel more comfortable (I've never used C#) and because I've already implemented the graphic interface of the game in this language.
So maybe I can collaborate with Kristján in this project (especially in the first phase, when we need to learn different techniques). Maybe we can implement different approaches for the same problem.
Of course if you (Kristján) want to participate, tell me.
|
|
|
Post by Þorgeir Karlsson on Feb 28, 2011 10:29:15 GMT -5
I have an idea for a program that would be able to predict football outcomes based on prior knowledge of the teams. The way I see it working would be a database of football matches from a certain league over the last year or two, and then when two teams meet up, the program would determine who is most likely to win based on how teams fair against each other under similar circumstances. It would then maybe be able to find patterns, such as if one team is especially good at winning away games, or if another team always draws when it plays a better team at home etc. There would be a lot of variables so an evaluation function would need to be created isolating the most important variables to consider, such as league position, current form, if the team is home or away and so on. What do you think Hannes? Does this idea have potential?
|
|
|
Post by Eiríkur Fannar Torfason on Feb 28, 2011 17:46:46 GMT -5
Grímur and I are interested in learning more about constraint satisfaction problems and scheduling. To that end we'd like to take a stab at creating a program that generates timetables for schools. There are several constraints that come into play in that particular real-world problem like room availability, teacher availability, room size vs number of students, equipment etc. When those constraints have been satisfied we can start looking at minimizing the number of conflicts for students so that as few students as possible have classes at the same time (assuming we know beforehand which students will be attending which classes). We might also want to look into grouping classes into sets which are not allowed to conflict (for example all classes for students at the same year in a given department).
|
|
|
Post by sindrib on Mar 1, 2011 8:59:00 GMT -5
My project is Windows application analyser, with the focus on testing methods that could identify/classify runtime functionality of the target application. The data used is retrieved from the process memory image at various stages of execution.
The project is written in C++, if someone needs a group, they are welcome to join.
|
|
|
Post by lorenzo on Mar 2, 2011 5:50:45 GMT -5
I want to join if it is possible to the project of Carmine the sudoku solver in java.
cheers,
Lorenzo
|
|
|
Post by Angelo Cafaro on Mar 2, 2011 6:59:06 GMT -5
I have an idea for a program that would be able to predict football outcomes based on prior knowledge of the teams. The way I see it working would be a database of football matches from a certain league over the last year or two, and then when two teams meet up, the program would determine who is most likely to win based on how teams fair against each other under similar circumstances. It would then maybe be able to find patterns, such as if one team is especially good at winning away games, or if another team always draws when it plays a better team at home etc. There would be a lot of variables so an evaluation function would need to be created isolating the most important variables to consider, such as league position, current form, if the team is home or away and so on. What do you think Hannes? Does this idea have potential? Hello! Yes this idea has potential, but in order to make this project a pure statistical analysis of some data you should include (as you said) several variables and maybe classify teams from different perspectives (i.e. one rough classification would be good teams, bad teams). In addition your program could learn from new results (not only using previous ones) and update his knowledge. Imagine you are building the "X league Master" that knows "everything" about the X league and can advise you on bets!
|
|
|
Post by Angelo Cafaro on Mar 2, 2011 7:01:11 GMT -5
Grímur and I are interested in learning more about constraint satisfaction problems and scheduling. To that end we'd like to take a stab at creating a program that generates timetables for schools. There are several constraints that come into play in that particular real-world problem like room availability, teacher availability, room size vs number of students, equipment etc. When those constraints have been satisfied we can start looking at minimizing the number of conflicts for students so that as few students as possible have classes at the same time (assuming we know beforehand which students will be attending which classes). We might also want to look into grouping classes into sets which are not allowed to conflict (for example all classes for students at the same year in a given department). Very interesting idea!
|
|
|
Post by Helgi Siemsen Sigurðarson on Mar 2, 2011 9:23:16 GMT -5
I would like to join Carmine for the sudoku solver in java, is the group full ?
Helgi Siemsen Siguðarson
|
|
|
Post by helgil08 on Mar 3, 2011 10:10:48 GMT -5
Hello.
I'd like to do a version of the Wumpus world in Java. There's already information regarding the problem in the material and it's an implementation of blind search with a knowledge base but I don't know if it's big enough to qualify as a final project.
Regards Helgi Leifsson
|
|
|
Post by krafki on Mar 4, 2011 5:05:21 GMT -5
Hello! The bot for action game sounded interesting and I ran into this Pogamut platform. It utilizes Unrealscript (Unreal Tournament 2004 scripting language) and also Netbeans for creating AI for UT2004. I was thinking about creating a deathmatch UT2004-bot using A* for path finding, and bot doing some basic actions when it comes across with an opponent. Link to Pogamut website: diana.ms.mff.cuni.cz/main/tiki-index.php?page=About
|
|