Assigned projects (for those who did not pick themselves...)

Proposed projects for Multi-Agent systems course

Cheating in auction protocols

        simulate the cheating in auction:

1. 5 participants bid for several goods in 4 kinds of auctions (english auction,
dutch auction, first price sealed bid auction, vickrey auction).

2. client want to use "take it or leave it" that make a deal with auctioneer in
advance, the auctioneer get some money and don't let the goods appear in the

3. for sealed bid auction, client can buy information of others, and put the
price a little bit higher than the highest price so that get the desire goods in
high profits.

4. for english auction, some of the clients try to form coalitions to keep
bidding price low. In particular, some agents will be avoid competition with
friends. some friends form a new agent, and compete with other agents.


1, what's the cheating mean? if all participants cheating, there is no cheating
in auction. In particular, if all agents can contact auctioneer to buy the
information of offer others bid, that not means cheating, that means open price
2, all the clients have a total amount of money, they also have their respective
perference for particular goods. In this case, I plan to design the perference
of the goods by using percentage. Total amount of money times percentage means
the evaluation of the price for the good.
3, I don't think the evaluation will be a fix price, instead I prefer the range
of price client want to pay. For english auction, they will not accept the price
higher they their own highest point of evaluation price. In dutch auction, they
prefer the decending price closing the lowest point. In sealed bid auction, they
perfer to give the price in the middle of the range.

Team members Yi Luo

Spectrum allocation

There is a rectangular area A where a spectrum owner is managing N=11 channels of spectrum. In each spectrum you have a total bandwidth of 50Mbit/sec and at most 30 nodes can have overlapping transmissions. A transmission range of a node is a circle with radius R(node) - depends on the node. Nodes can buy the right to transmit from the spectrum owner - depending on the node they want to communicate with. Sometimes the owner might want to buy back bandwidth from the nodes, and potentially give them more in other bands.

Team members Nicolaos Peppas, Mohammad Ahmad Zubair, Haroon Shariff

Target tracking in sensor networks

You need to buy time and antenna direction to track a moving target in a field of sensors with directional antennas. You compete with other participants.

Team members: TBD


Simulate the interactions in the reality TV show "Survivor" in an agent oriented environment. Experiment with different strategies.

Random thoughs: you can approach this as a multiplayer game theoretic problem - a relatively complex game, with sometimes counterintuitive rules (the fittest not always wins, etc.). Another idea is to model this as a genetic algorithm problem - and try to develop a way to automatically generate strategies.

As a group project, I would expect some students to take care of the communication framework, voting out process, environment, etc, and some students to deal with the strategies of individuals.

Team members: TBD

Unmanned Aerial Vehicles

The project will consist of a number of unmannel aerial vehicles (UAVs). The UAVs will coordinate with one another in order to meet mission parameters. Mission parameters will be set forth by an end user at a console, but user participation can be minimal. Mission parameters may include target acquisition and tracking, attack, and communication relay. The UAVs will be simulated based on real-world aircraft flight characteristics. The UAVs will communicate using a message-passing protocol over serial radio with limited range. The user interface will consist of a map displaying a United States Geographical Survey (USGS) map and the constantly updating relative positions of the deployed UAVs. The map will also display approximate location of targets and other areas of interest identified by the UAVs.

Team members TBD

Surveillance with agents

Using the Webots environment to create surveillance agents. The agents would do surveillance in a building, being able to map its environment using probability maps. They would somehow figure out a way to cover as much of a room or building as possible. If any unwanted activity occur, they would try to locate the source.

Team members: Joakim Ekblad

Trading agent competition

Create an agent for the Trading Agent Competition Trading Agent Competition environment. We chose the TAC-SCM game using the AgentWare API interface that is provided by the organizers.

Team members: Brent Horine, Ghaith Hadddad

Agents in a game environment

I will be creating an agent in a game enviornment (Quake 2, Quake 3 Arena, or Unreal Tournament), although the exact game is still under consideration. The agent will be modeled with Context Based Reasoning, and despite our previous email, I will not be implementing this using SOAR. The agent will be actually be an A.I. bot inside of the game. The bot will be an opponent for a human user to play against.

This is related to my thesis research, which in short, is to implement and test machine learning techniques with context-based reasoning in a game enviornment. Basically, this will be the first part of the research, which is to implement an agent in a game enviornment using context based reasoning and test it against human players.

The deliverables for this project will be the source code and any other files that will be needed for you to load the agent (bot) into the corresponding game. Along with the code, I will have a website which I will be posting my bot for users to download test and give feedback on. This feedback and data that will be collected from the website will also be summarized in the required report.

Team members: Chris Moriarty

Reliable and Redundant Grid Data Backup Using Agents

As disk space sizes and Internet bandwidth both increase for modern PC?s, they become more and more capable of offering resources in a computing grid paradigm. Spare disk space and CPU power are common on these machines, and could be exploited for other users while idle. Also as computers become more important in the everyday lives of people, there is a greater necessity for reliability backing up or preserving the integrity of the user data. In this project, a agent based system is proposed as a solution to this difficulty. Each user that is part of this ?Confederated Backup Network? will share part of their available disk space and processor with the network. The PC will spawn an agent that will be responsible for managing the user?s needs. This includes brokering with other PC?s agents to exchange blocks of user data, to allow for greater redundancy. The role of the agents will also include bargaining for the necessary backup space with whatever currency it is authorized to user (i.e. it?s own space for redundant backup, CPU power, or money). In this way, the user will be able to specify an amount of backup storage with an arbitrary reliability, and have the agent broker for the user?s needs.

Team members: Jimmy Secretan, Malachi Lawson

Agent for real estate
Description: My project idea is that of a real estate agent. This agent will take some parameters from the buyer and search for possible matches and communicate them to the buyer. They will have a communication framework, where a buyer agent and a seller agent could communicate and negotiate, and then present their solution to the human buyer and seller for sign off and then creation of a contract.
Team members: Asher DeVuyst