Categories and Subject Descriptors
1. INTRODUCTION 1http://www.microsoft.com/elearn/support.asp 2http://www.peer3.com/text/software/software.html 3http://home.click2learn.com/products opment can be also extended to resources or metadata secthe tableofcontents structure of the organizations section in the manifest le. Nevertheless, collaborative develtions. resources, and metadata). Then, the author submits a section 3.3. Then, a negotiation protocol is executed by continues until an agreement is reached or some degree of substituted by a further elaborated proposal. This process proposal, including the dierences bet ween both les, and refering to tableofcontents negotiation. The rest of collabconsensus is achieved. orating authors receive and evaluate the proposal, according to a set of previously agreed criteria, like those described in every author until the proposal is eventually accepted, or 1.2 Cooperative development of learning objects mains. multiple-tier architecture, where cooperative agents collaborate in lower-level domains to consolidate some knowledge, architecture. Nevertheless, it can be easily extended to a Because of simplicity, we dene a t wo-tier cooperative work before they try to collaborate or compete in higher-level do2. COOPERATIVE KNOWLEDGE PRODUC
TION ARCHITECTURE ticipants may keep their own pace within the interchange. ulty sta in a virtual univ ersity, cooperativ e development In a highly distributed environment for instance, the facAlthough, a group of agents can jointly, asynchronously dequent. The exchange of ideas between members of the dischronous ph ysical or virtual meetings is quite less fretivities. of learning objects becomes harder, since the holding of synvelop learning objects if they coordinate their creation actributed workgroup is an asynchronous process, where parCooperative Knowledge Mart
Supervisor
Agent
Cooperative Knowledge Mart Supervisor
Agent
Cooperative Knowledge Warehouse
Knowledge Mart C 3. AGENT NEGOTIATION IN A KNOWL
EDGE MART thus recording the history of negotiation and their decisions. 3. Agents can deliberate and achieve an internal state, 3.3 Proposal Relevance
3.1 Negotiation Principles
3.2 Messages 4. NEGOTIATION PROTOCOL IN A CKM dependent on attributes of k, in such a way that if of a proposal k in a CKM returns a numerical value, 6= then 6= ki kj, u(ki) u(kj). Proposal relevance function The relevance function u(k) Rule 3 If receives a n) message from other Ai proposal(kj; Then, proceeds as follows: Ai uates the new proposal . If then sets a kj ki kj, Ai new time out waiting for proposal to be ratied. t1, kj agent referring to the same negotiation n, eval- Aj, Ai 1. If does not receive any proposal referred to Ai negotiation n before expires, then initiates t1 Ai back the protocol in Rule 1 with the same proposal ki. same negotiation n, then gives up the initial Ai 2. If receives a n), with Ai consolidate(kj; kj ki, proposal and the protocol nishes unsuccessfully . for j 6= i, before expires, and referring to the t1 phase receives a proposal that is evaluated as (t0-waiting) the distribution phase, that occurs after an agent submits a preferred. that can be noticed in the negotiation process. is used for T0 tion phase, that occurs if an agent that is in its distribution We are considering two dieren t timeouts, one for each phase proposal by executing Rule 1. is used for the consolida- T1
Rule C If did not sent any message referred to the Ai same negotiation n, it does nothing. in the negotiation process back to and Aj, extends timeout. t1 If then sends last proposal it sent kj ki, Ai following actions: Rule A If sent a n) and is waiting Ai proposal(ki; for timeout, then it can perform one of the t0 If receives a ; n) from Ai consolidate(kj Aj:
4.1 Activation Events processes. Each negotiation process is handled separately, by initiating a new execution thread of the protocol.
A cooperative agent can participate in several negotiation Ai ously submitted proposal, until a counter-proposal is elabposal that is evaluated as preferred, a new timeout is set t1 waiting for timeout. Then, if the agent receives a pro- t0 orated. An agent that has not reached this state will be to give it a chance. But if the preferred proposal is not eventually ratied, then the agent goes on about its aims agent may provoke a momentary retraction from a previand will try again to consolidate its own proposal.
In any moment, the reception of a message from another with the same proposal. as in step 3-1 and initiates back the protocol If then acts in the same manner kj ki, Ai 4.2 Protocol Variants 4.3 Implementation are proposed to be dynamic, depending on agents’ ontologytiveness of the model. Our work does not consider yet how The architecture presented here is a bottom-up approach to well-dened. The structuring of knowledge marts can help holds responsibilities on some domain level knowledge, while cooperation level knowledge interfaces to other domains are to other knowledge production tasks, as software developbased expressed interests. knowledge marts are set up. As a future work, knowledge further validation is needed to assess the usefulness of the to reduce inconsistencies between agent territories. The coment, specially in analysis and design phases. Nevertheless, the design of cooperative multi-agent systems. Every CKM mart generation and the participation of agents in CKMs protocol in dieren t scenarios, and we are conducting tests operative approach presented in this work is also applicable on the impact of the number of agents in the overall eec