CSci 8551 Fall 2017: Intelligent Agents

Project Requirements

For your class project you can choose any topic related to the material covered in class. The course project is meant to give you a deeper "hands-on" understanding of the material. During the course you will come across some ideas that you will find interesting, and you will be interested in pursuing them.
The project should be at the frontier of research in agents, but it should not necessarily be innovative. Duplicating research results that you learned in class, extending them to address some limitations, or examining their strengths and weaknesses on different problems are all good ways to do your project. This is not to prevent you from doing something really innovative!

The project can be used as a 50 hours project for Plan C. If you want to use it for Plan C, it has to be done individually. If not, you can work with one other student to the same project.

The project can be an experimental project, where you implement an agent system and do an empirical study on its performance, or a theoretical project, where you present theoretical results and support them with proofs.

Project Due dates

The project is due in three written parts and an oral presentation:

The first written part is a preliminary topic title, author(s), and a few references on related papers. It is due on October 25 .
The second part is a progress report. It should be 3-4 pages long. It is due on November 15 . It should include:
- a description of what you have accomplished so far
- preliminary results (if you install existing software show that it works properly on a few examples, etc.),
- a plan for the remaining work.
The last part is the final report. It should be 10-12 pages long. It is due on December 16 .

The report for the project should include a description of the problem and its relevance (1-2 pages), a short overview of related work (1-2 pages page), a description of your approach (1-2 pages), results and analysis of your work (4-5 pages), conclusions and future work (1 page), and references (1 page). You do not need to include source code, but you should attach traces of runs in an Appendix.

Project Presentations

Oral presentations will be scheduled for the last 4 class meetings.

There will be 4-5 presentations per lecture.
Each project will be allocated 10 minutes. Projects with more than one student will be given some additional time. The time includes the presentation time (6-7 minutes) and time for questions (3-4 minutes). This means you should not try to explain in detail the entire project. You'll have just enough time to summarize the topic, the methods used, and the results.

Grading

The 35% score for the project will be distributed as ffollows:

5% for the initial submission.
5% for the second part.
20% for the final report. The report will be evaluated as follows.
- 7% quality of the report
- 3% complexity of the study
- 10% analysis and evaluation of results
5% for the presentation in class. The presentation will be evaluated on length, quality of presentation, quality of visual materials, quality of answers to questions, and ability to stimulate thinking.

Project Resources

Look at public domain software systems for sources of information about software for agents.

Titles of past projects

Multiagent, Frontier-Based Exploration in a Simulated Two-Dimensional Environment
Implantable Medical Devices as Agents and Part of Multiagent Systems
Extensions on Autonomous Distributed Construction Methods
Implementing antlike exploration behavior in a small set of mobile robots using Player/Stage
Effects of Real time bidder support in Combinatorial Auctions
Performance Evaluation Methods for the Trading Agent Competition
The Cause Implies the Cure: A Simulation
Security framework for Vickrey auctions for telephone or VoIP calls
Modeling Comment-Spamming Agents
Approaches to RoboCup Keep-away Soccer
Rule-Based Behavior and (American) Football
Bidding agents for Combinatorial Auctions
Robotic Swarm Dispersion Using Wireless Intensity Signals

Ideas for projects

More to be added.

use StarLogo or Swarm to model and study the behavior of a simple decentralized system.
develop a communication language for robots to communicate to each other and share information.
develop a market-based approach to task partitioning. Design algorithms and test them for a specific task or task domain.
design algorithms and strategies for partitioning a task among multiple cooperating robots. Implement and test them on a specific task.
install and evaluate a public-domain package for building agents and use it to develop an application.
install two or three public-domain packages for building agents and compare them. Issues to consider are flexibility, robustness, ease of use, documentation available, etc.

Avoiding Plagiarism

Plagiarism definitions will help you understanding if you are plagiarizing others work in your report. Take a look at it.

Typesetting your paper

You should learn Latex for typesetting your paper. It will take some extra time, but the effort will pay off in the long term.
Here you can find Information and pointers on Latex.