University of Calgary
Rob Kremer
Group Assignment 4: Auction Protocols in CASA

CPSC 601.68/599.68: Agent Communication
Winter 2007
Department of Computer Science
Computer
Science

Note: This is an group assignment. The work is shared among the group members and each member is responsible for all of the work. For example, if there is plagiarism is found in the assignment, every member is responsible. The group is expected to do it's own work, and must properly cite any and all work taken from external sources.
Each member is responsible to carry their own portion of the workload. To this end, there is a peer assessment process where each member must assess each other member's contribution. Individual marks may differ based on the peer assessment, the instructor's assessment, and other factors.

Each group will choose a different auction type from the following list:

You can find details of the auction types at (Reynolds, 1996). The chosen auction will be implemented in CASA.

The idea behind the assignment is to run an auction with one seller and at least 3 types of bidders (eg: "aggressive", "conservative", "middle-of-the-road", "risk-taker", "randomly-risky", "pure-random", etc). You will concoct an experiment to test which if the bidder types works best for your auction type. To do this you should run successive auctions, say, auctioning off 100 items in 3 different categories (A,B,C) where item has a (secret) value, but where the values of each item category are normally distributed normally around a publicly known value. The actual value of each item will be sent from the seller to the successful bidder as part of the notification of success. If each bidder starts with a set amount in it's "bank account", the winner could be determined by the "most money" at the end. The winner bidder type can be determined by the average bank account balance for the type (you may want to consider additional penalties for 0 balances, etc). The details will vary, of course, depending on your auction type.

Example experimental parameters might be:

Starting banking accounts: $100
Bidder types: aggressive, conservative, random
Number of bidders of each type: 10
Item types: A, B, C
Number of items of each type: 35
Item price distribution: normal distribution, SD is 30% of value, A mean: $10, B mean: $50, C mean: $100

You are free to modify the parameters, and even the experimental setup to suit your particular auction type and experiment. You are advised to discuss this with the instructor to head-off any problems in the demos.

Marks will be based on (but not limited to):

The assessment will based both on a one hour demo (to be scheduled), and on handed in code (pack it in a .jar file [include .java files] and mail it to the instructor).

References

Kate Reynolds. Auctions: Goint, Going, Gone! A Survey of Auction Types. Agorics Inc. Los Altos, CA, USA, http://www.agorics.com/Library/auctions.html. 1996.


UofC
CPSC 601.68/599.68: Agent Communication
Department of Computer Science

Last updated 2007-04-02 9:36
Rob Kremer