CPE/CSC 580
Artificial Intelligence -- Intelligent Agents
Spring 2006
Status |
Final |
Points | 10 |
Deadline | Tue, May 23, 23:59:59 |
CPE/CSC 580Spring 2006 Artificial Intelligence -- Intelligent Agents
Assignment 2: Agent Demo
Goal
This assignment continues the exploration of agent development environments
through the implementation of a sample agent (or team of agents) that demonstrate
the capabilities of the environment.
Description
For this homework, you need to identify a suitable task for demonstrating the features
of your selected environment. You can either select the same environment that you examined
for the first assignment, or one of the other ones discussed by somebody else.
Agent Capabilities
The agent should demonstrate at least three of the following capabilities:
- Autonomy: Is the agent capable of
goal-based problem solving without significant interaction?
- Mobility: For a software-based agent, this typically requires
state preservation,
packaging of the agent (e.g. serialization)
transportation through use of existing protocols and infrastructure (e.g. http, rcp)
unpacking, and
resuming its operation at the destination.
Since the infrastructure for mobile software agent is not very well developed,
you could possibly include visual display of movement of the agent here
(e.g. games, artificial life).
If appropriate, you can also consider agents that remain stationary, but have a long "reach"
(e.g. Web spiders), and thus accomplish similar tasks as mobile agents, only with different means.
- Knowledge Representation and Reasoning:
Does the agent have an explicitly represented internal world model, and is it capable of
drawing inferences based on the world model?
- Planning: Does the agent perform the
generation and execution of (partial) sequences of actions to accomplish a task?
- Learning:
Does the agent achieve performance improvement through adaptation of functional components or knowledge?
- Multi-Agent Tasks:
Are multiple agents involved in resource sharing, collaboration, or the
coordination of activities (e.g. collision avoidance) in order to achieve
a common goal (or at least compatible individual goals)?
- Communication:
Are the agents involved in the exchange of meaningful information?
Do they use an agent communication language, speech acts, or conversation protocols?
Submission and Deadline
Your material should be submitted to the respective Blackboard discussion forum.
Please create a thread as soon as you have selected the task and environment for your agent(s),
and then post your material.
Grading Criteria
The number of points for this homework is 25.
The score will be determined by the following criteria:
- knowledge of subject matter
(does the author demonstrate a good grasp of the topic)
- evidence,
(does the author present facts, examples, or other material
that supports the arguments, and are the sources cited)
- complexity of the tool investigated,
(what is the scope of functionality available through the tool)
- thoroughness of the evaluation,
(are all relevant features of the tool examined)
- accuracy
(is the material presented correct, consistent, and complete)
- organization and readability of the write-up,
(is the paper well organized and easy to follow)
- presentation
(is the appearence of the paper of professional quality,
e.g. without spelling or grammatical errors)
A tentative version of the evaluation form I will use is available here:
580-A2-Evaluation.txt;
it is essentially the same as for the first assignment.
For a high score, your evaluation should probably be at least three pages long.
You can use the outline below as guidance for the structure of your writeup.
Identify and discuss the main characteristics of the task that your agent
(or team of agents) will perform.
Select the capabilities that your agent will be endowed with, and describe
how it will use them to accomplish the task identified above.
Describe the overall structure of your agent, and how this embodies
the capabilities dicussed in the previous section.
Discuss the strategies, methods, and techniques used to implement
your agent in the selected environment.
Perform an assessment of how well the agent performs its task.
This can range from a proof of concept that simply demonstrates the feasibility
of a task with a given agent design, to specific performance measurements.