Thursday, 27 November 2008

Making the implementations portable

I realised yesterday that my implementations (of the eumas- and aamas- negotiation policies) are written such that they depend on the particular setup/configurations of my office computer and may not work elsewhere. The main problem is the way Sicstus Prolog is called from the Java (Jade) code. I think I have solved the problem such that my code should run on any Linux machine. However, as for Windows... having a few problems!

General argument evaluation procedure

I completed last week a general argument evaluation procedure (written in Prolog) that given a claim in the context of an Assumption-Based Argumentation (ABA) Framework (consisting of a language, inference rules, assumptions and contraries) checks whether the claim is acceptable according to the admissibility semantics and, if so, returns the defence set (which includes facts as well as assumptions).

I will now modify the implementations of the eumas- and aamas- negotiation policies to use (call) this procedure rather than CaSAPI.

The reason for implementing another argument evaluation procedure when CaSAPI already exists is that CaSAPI contains many features which I do not need and does not contain some features which I do need (e.g. returning a defence set that includes facts, as required in the kind of inter-agent communication setting I consider).

UCL-Imperial Workshop

Took part in a workshop organised by Francesca Toni (Imperial College) and Anthony Hunter (UCL) on Tuesday. The aim was for students (including me) to present and discuss work in an informal setting and to receive feedback. I have uploaded the slides I used to my home page.

Monday, 17 November 2008

AAMAS paper implementation

Completed implementation of the policies described in the AAMAS 09 paper, modifying the paper along the way and not replicating the policies in the paper entirely. Also identified a few problem scenarios (see 'readme') that demonstrate why the policies are not 'complete'.

Plan now is to define a general argument evaluation procedure before proceeding with anything else.

Blocking Initiator Behaviour

Modified the Initiator Behaviour (for the EUMAS and AAMAS paper implementations) so that it does not 'block'. Example 5 of the EUMAS paper demonstrates the problem: The initiating condition test does not succeed first time round and blocking at this point could mean that the Initiator Behaviour is not scheduled again.

Ordering of requesting, responding and receiving a response

According to my current implementation, agents check that there are no incoming messages (requests or responses) before sending/initiating a request. That's fine. The problem scenario occurs when an agent receives a response and a request. Does the ordering of which it processes first matter? Yes, I think so. The agent should process any incoming responses (updating its belief base as necessary) first before responding to a request.

Submitted EUMAS paper

Completed revisions of the EUMAS paper ('On the benefits of argumentation - preliminary version') and submitted the revised version. Also, prepared an extended technical report and sent it off for uploading to the Department of Computing webpage.

Monday, 10 November 2008

Speaking Technically

Read this book ('Speaking Technically: Handbook for Scientists, Engineers and Physicians on How to Improve Technical Presentations' by Sinclair Goodland) whilst preparing my EUMAS presentation. Quite useful. A good book to refer back to in future.

Thursday, 6 November 2008

EUMAS paper revision

Had my EUMAS paper accepted, al7amdulillah. Spent this week making modifications based on reviewer comments and also working on the slides for the presentation in December.

AAMAS paper implementation

Began implementing the negotiation policies of the AAMAS paper last week. Completed the simple policy, albeit by means of a makeshift argument evaluation procedure. Still need to do the argument-based policy, including the updateBeliefs procedure.