CUHK CSC7112 Blog

Agent Development and Implementation

2009-02-27T01:34:00.000-08:00

As requested by a student, I would like to introduce to you a commonly used Java platform for agents is JADE:
http://jade.tilab.com/

It supports ontology in its form. See this tutorial:
http://www.ryerson.ca/~dgrimsha/courses/cps720/JADEOntology.html

Explicit OWL/RDF support for JADE is available:
http://agentowl.sourceforge.net/

which is based on Jena:
http://jena.sourceforge.net/

Protege has a plugin for generating Java files representing an ontology that can be used with the environment:
http://protege.cim3.net/cgi-bin/wiki.pl?OntologyBeanGenerator

More on your survey paper.

2009-02-18T18:48:00.000-08:00

After viewing a number of your drafts, one main point I have to stress is that you need to include quite a number of related work citation and references, which is related to your topic. You may search in Google Scholar for something like "XX agents", "XX ontology", "XX intelligent", "XX planning", "XX negotiation", etc. At this stage, you should not worry about the solution but more survey what others are doing related to your topic.

To access major digital libraries outside campus, you can first install the VPN of our department. See: http://wiki.cse.cuhk.edu.hk/tech/network/vpn

Then set your web browser proxy to: proxy.cse.cuhk.edu.hk (port 8000)

I have posted this 2 links of tutorials for writing survey papers in the course webpage:
http://iaser.tek.bth.se/feldt/courses/pa2403/pa2403_lecture2_070906.pdf
http://www.eecg.toronto.edu/~bli/grad/lectures/Episode%202.pdf

If the class indeed need one more week, pls vote for this in class.

Lecture 5 - Ontology Standards RDF and OWL

2009-02-18T07:52:00.001-08:00

The ontology standards RDF and OWL have rooted from AI and therefore has a strongly AI favor. Some interesting flexibility of RDF like reification, blur distinction between attributes and relations can be dated back to old semantic net in AI.

You may compare with the UML modeling, which I suppose most of you know.
Even UML supports the Object Constraint Language (OCL).
We can see platforms and paradigms are therefore moving towards one another.
RDF does not have some typical modeling constructs of UML like cardinality, etc. So, OWL adds those in.

Although detailed syntax will not be examined, you should still know how some of those key features may be used with agents and intelligent system integration.

To show ontology examples, probably it is much more convinent to use diagrams, which are widely supported by many tools (or just use UML notations that you are probably familiar with).
A well-known free tool is called Protege. See: http://protege.stanford.edu/

Lecture 4 insertion: agents and alerts

2009-02-05T21:58:00.000-08:00

As requested and pointed out in the last lecture, I shall insert the talk about "Agents and Alert Management" to help link your papers to my research foundations.

Again, don't worry, this is a practical design methodology, not a theoretical one.
But of course if you are interested in theoretical research on this topic, like linking game theory to alert manangement / workforce allocation, simulations on allocation strategies, etc., you are always welcome and I can fully support your work :-)

Other practical topics, like semantic based alert management with agents are also useful and interesting.

Survey Paper deadline extended with some suggested topics

2009-02-05T21:56:00.000-08:00

Draft survey paper due Feb 15, survey paper submission due Feb 22, 24:00 (extended).

*** After viewing a number of your drafts, one main point I have to stress is that you need to include quite a number of related work citation and references, which is related to your topic.

You may search in Google Scholar for something like "XX agents", "XX ontology", "XX intelligent", "XX planning", "XX negotiation", etc.

At this stage, you should not worry about the solution but more survey what others are doing related to your topic.

Please discuss your topic and have my approval. If you have no idea on your topic yet. Please consider:
agent based workforce allocation
agent based urgency management
agent based supply chain management / logistics management / service composition
agent based security management
RFID, smart objects, and agents

Survey paper guideline

2009-01-20T10:15:00.000-08:00

The guideline for the survey paper is now available. (htm)
Please choose your topic early as the due date is in mid-February (Feb 15, 24:00 ).
Students are request to sent in the following information to discuss your topic.
I will give you hints based on your profile.

Please tell me the following about you in your email:

. Undergraduate / other degree (MBA / MSc, etc.) program:

. Undergraduate / current / other degree projects:

. Current MSc program and course taken:

. Current / Previous Job (job function and business sector):

. email address (prefer "permant" one that can contact you after graduation)

. mobile phone (optional)

Lecture 3 and 5

2009-01-20T10:06:00.001-08:00

We shall continue to use negotiation as an example to illustrate how ontologies and semantics help in MAS. Besides helping the agent to understand concepts, ontologies and semantics also help humans to understand their requirements and therefore help the specification of requirements and preferences better.

For example, in Yahoo auction, can you easily find a digital camera with the requirement of "better than 3 megapixals" and "fitting CF or SD card"? With ontologies, we can.

After the motivating example, we shall then go over some technical details of RDF and OWL, with nowadays are essential for many other "intelligent" or searching applications, not just for agents.

Lecture 2

2009-01-12T23:12:00.000-08:00

One of you commented the concepts of agents are still fuzzy for them.
So, in Lecture 2, we introduced some internal mechanisms and architecture for agents, especially the Belief-Desire-Intension agent architecture.

We showed a detail example of a concrete BDI agent the paper on negotiation based on constraints as to illustrate the functioning of agents:

D.K.W. Chiu, S.C. Cheung, P.C.K. Hung, and H.F. Leung. Constraint-based Negotiation in a Multi-Agent Information System with Multiple Platform Support, 37th Hawaii International Conference on System Sciences (HICSS37), Jan 2004. (ppt)(doc)

We shall go over the rest of the paper in the next lecture.

So far, we have introduced both the macroscopic (top down) and microscopic (bottom up) view of agents and MAS.

Lecture 1

2009-01-12T22:44:00.000-08:00

We have introduced what are agents and what are Multi-agent Systems (MAS).
We have also introduced a motivating case study that uses agents as software building blocks
for a service-oriented (telecom) enterprise. This serves also as the vision what the ultimate product that we aim at in the (near) future, because now we have the enabling technologies ready, together with the arising need of business intelligence from various businesses. If we were to use classical examples of agents in robotics and industrial control, those are probably too far away from your experience and work. We shall refer back to this case from time to time for discussion.

One of you asked the difference between Web Services and agents. Similiar to Web services, agents not just serve as software building blocks but also as integration mechanism, having interfaces and protocols. However, agents means more:

Web services are usually just reactive (on invocation), but not proactive
Web services are normally for individual invocations like functions and procedures, but agents normally run continuously as a long-life process to help its owner / delegator (e.g., appointment agents), i.e., autonomous
Agents can include intelligence and flexible actions (especially exception handling), which is not common in current Web services. (flexibility)

As mentioned in this paper, this is a foundation for future detailed design, analysis, and research. So, if you are interested in any part of this (say, capability matching, location-based allocation), you are encourages to choose one of this as your survey (horizontal / in-depth approach) as well as your course topic.

Alternatively, you may also choose another domain of your choice, usually integrating knowledge from your job (vertical approach), your degree project, or your favorite courses.

We emphasize on practical relevance. So, please make sure that you are able to give examples to illustrate your work.

Welcome to our Course!

2009-01-08T10:49:00.000-08:00

The course Webpage is at: http://www.cse.cuhk.edu.hk/~kwchiu/csc7112/