Labels

Showing posts with label MAS. Show all posts
Showing posts with label MAS. Show all posts

Tuesday, November 8, 2011

Feedback on MAS2

Tijmen,

In principle I don't provide feedback on assignment 2, but if you'd like feedback drop me a line.

You're right about assignment 1; I updated your grade on Blackboard. Your final grade is not affected.

Cheers
/``````````````````````````````````````````````````````````````````````````````````
MASS spam Matthijs' mailbox now with requests on feedback! :P

Tuesday, November 1, 2011

Pursuit domain exercises: post-mortem

How did everyone like the MAS assignments?

I thought they were pretty nice. I especially liked how you were allowed to experiment a bit and figure stuff on your own. I enjoyed coming up with wacky algorithms or wacky improvements/variations to given algorithms.

I didn't like the competitive element (but maybe that's because I don't like losing? :P) with the capture times and whatnot. And I really didn't like debugging technicalities that had more to do with the limitations of the package we had to build on than anything else.

Anyway, I decided to make all my code publicly available. I used version control, so you can even see what improvements I made in each version (also, how I left both assignments for the last minute). I also uploaded both reports, which you can find in the downloads section. Here's the main page:

Sunday, October 30, 2011

MAS assignment 2 deadline

There was some confusion over the deadline for the second MAS assignment, about whether it was final or points get deducted if you are late. I sent an email to Matthijs and he just replied: 
0.5 grade point per day
So, to re-iterate, assignment 2 is just like assignment 1; you get a penalty of half a point for each day after the deadline you are late. So, if anyone feels that one more day can really make a difference (like me), then this is good news.

Saturday, October 29, 2011

Designing an unbiased reward function

Still doing part 1 of the second MAS assignment. I've been wondering about the proper way to set the reward function. If we want our system to learn, the reward function should only include the objective of the game, right? If we start letting it know that collisions are bad or that moving closer to the prey is good, then what's the point of using Q-learning? You might as well just give it the perfect algorithm from the get go.

Is my thinking right? If it is, then the correct way to do this is to only give out reward when the prey is caught. Then, the Q-values should propagate the information. For example, since collisions would delay this reward, then the Q-values of moves that are likely to lead to collisions would become low in the long term. Does that make sense?

Tuesday, October 25, 2011

MAS exam "bootleg"

Hi everybody!

Since I got that we were not allowed to take the exam questions with us I toke pictures of them:

https://docs.google.com/open?id=0BykUHOwlm7yaYmI0YzFhZWQtY2ZmMS00ZWNkLWFmOGEtN2NhNzBlMzY4OWRi

Enjoy :-)
Dario

P.S. Have you red about John McCarthy? It's a sad period for the computer science...

Saturday, October 1, 2011

No-life programming on predators

Hey all!

Let's post us some vids of our working predators! XD. This is the one I made:

youtube video!

Averages at about 21.5. How did you guys do?  *curious* Apparantly the guys from last year had a record of 19.8 :(, did anyone get close to that? how? :O

Tijmen

Wednesday, September 14, 2011

MAS Slides

Hi guys,

as I've experienced some problems downloading lectures' slides for the Multi-Agent Systems course (poorly coded Adobe plugins for Linux), here I post direct links to them (probably you need to be authenticated into blackboard to access the files, but I'm not sure..):
  • Chapter 1.pdf (http://blackboard.uva.nl/bbcswebdav/xid-1356811_1)
  • chapter3.pdf (http://blackboard.uva.nl/bbcswebdav/xid-1356812_1)
Cheers,
Dario

Tuesday, September 13, 2011

MAS Predator Prey Assignments

I reckon we'll have to make a post for all assignments to get things going. Can anyone help me with this as I still do not have access to blackboard? XD

Anyways, link to the predator prey problems are here:
https://sites.google.com/site/matthijssnel/teaching/mas2011

The core of the assignment is the second part I think, where you have to coordinate similar agents to surround a prey to capture it. A few things to note already:

- Preys and Predators can occupy the same space, but preys and preys, as well as predators and predators can not (predators get warped away randomly... this could actually be used to your advantage perhaps? Capture 1 prey, move to same location, everybody warps randomly away :P)
- Preys don not actually move to squares that are occupied, so you can herd and direct them in a sense. Do not know if this will be useful or not though :)

Also, here are a few articles describing the problem to perhaps give you some ideas!:
http://rua.ua.es/dspace/bitstream/10045/8320/1/reverte08extending.pdf
http://rua.ua.es/dspace/bitstream/10045/8321/1/reverte08cooperation.pdf

As always, post your comments and questions and stuff here. I might upload some stuff on these problems later.

Cheers,
Tijmen

Friday, September 9, 2011

Guide to installing the MAS pursuit software on windows

Hey guys,

For the Multi-agent software course we have to do the assignments on the following webpage with the pursuit software

https://sites.google.com/site/matthijssnel/teaching/mas2011

I reckon most of you have downloaded the software, and go it running on the linux computers by now. If you only have windows at home and don't want to dual boot you can do the following to install linux within windows on your computer, so you can get the pursuit software running:

1. Download and install virtualbox
2. Download the Ubuntu  .iso file
3. Download daemon tools light
4. Daemon tools is used to use .iso files (which are actually cd/dvd like files) from you computer. Basically you fake your computer into thinking you have inserted a cd/dvd with the .iso file in your computer. You need to go to the tray with the daemon tools icon, right click it, go to virtual devices and add a new device. After that click on the new device in the same menu, and click on 'mount image'. Choose the path that leads to the ubuntu.iso file, and you are set to go.
5. Run virtualbox. This program will use some of your harddisk to install linux on, you can then run it via windows. In the software click on 'new', you will be taken to a wizard. Choose a name for your linux installation, and select Linux and Ubuntu. Select RAM to allot to linux, 512 will be fine for our predator-prey application. Choose 'create a new hard disk'. Select VDI, then dynamically allocated (unless you will be using it often, then choose fixed size), choose a location name and alot about 8GB to the linux OS (4.5 GB for the linux installation, + some extra space. 8 GB should easily suffice if you only use it for programming tasks like this. Press Create and you are done.
6. Now that you have created your linux virtual machine, you have to install linux on it. In virtualbox, click the 'start' button, and proceed with all the steps to install linux on your machine. Note: linux only has access to the 8GB of empty space you allocate it, so dont worry about the program deleting your files. Just choose: erase disk and install linux.
7. Now you have a working linux distribution! You can download the predator prey software to your linux OS. When you try to run the software you will most likely be prompted to install two packages. You can do this by going to System -> administration -> synaptic package manager. You can install the packages from here with a search (I forgot their names, but there was one tricky when where you needed version 5 instead of the 6 that was installed). For those that will run the java predators, you will also need to install the javaconmpiler from here. Either the sunSDK one, or openjvc will do.
8. Now that you have the packages, the script files (.sh, or shell) will be incomplete. Change any paths going to something.../bin to ../../bin   The ../ indicates the previous directory.
9. Everything should work now, compile the predators, run the .sh file and test it out!!

If something doesn't work for you, please let me know.

Tijmen