The Study, Based On Acts Of Self-confidence Of Agent Cooperation And Task Solving Mechanism

Because CSCW adapts the characters of work modes in the information society, such as group, interaction, distributivity, cooperativity. It has an amplitudinous application market with greate potential. And it shows a strong tendency to improve. At the same time, with the improvement of the technologies of computer networking and computer communication , the research on software Agent and multi-Agent system has been the focus of distributed artificial intelligence. The core of the conception of Agent is describing a dynamic entity, which can sdudy and improve by itself, and adjust itself to the environment. So Agent can incarnate the characters of actions in the collaboration team. In the multi-Agent system, one of the most important contents of the research on Agent is the study on how to collaborate among Agents, so that Agents can complete a task together. Therefore Agent technology can solve the questions of collaboration in the CSCW system. And with the Agent technology, we can overcome the shortcomings in the former distributed CSCW system. And we will benefit from using the technology to the collaboration in CSCW system.This paper lays emphasis on how to collaborate among Agents on completing a task in the CSCW system. The paper introduce a cooperative question solving mechanism which is based on the concept of self-confidence degree of Agent coalition's behaviors in completing a task. Firstly, the concept of Agent coalition is introduced. Secondly, self-confidence degree of Agent coalition's behaviors in completing a task is introduced, and its calculation expression is given. Some modification rules for self-confidence degree of Agent coalition's behaviors in completing a task which Agent coalitions should meet in multi-Agent coalition system are also characterized. Finally, a task entrusting and executing algorithm is designed based on the theory of self-confidence degree of an Agent coalition's behaviors in completing a task. In this paper, based on the task partition AND/OR tree, the self-confidence degree of a single Agent and an Agent coalition's behaviorsin completing a task is defined. This is better than the concept of self-confidence degree based on the activities in the reference [1], because it is easier to be operated and grasped. Using abilities and action modes to estimate the self-confidence degree of an Agent or an Agent coalition's behaviors in completing a task, the ability in completing a task is better showed compared with reference [1]. In this paper, the rules for balancing the load, estimating risks and blacklists is introduced, which meets the actual needs of competing question solving. The rules for modifying the self-confidence degree when an Agent has joined into an Agent coalition and an Agent is divorced from an Agent coalition , which meets the needs of the mechanism of forming Agent coalitions. In addition, the safety intimidation the multi-Agent coalition system faces is analyze , and a safety controlling mechanism is designed.In the realization part, a framework of an Agent's activities body is given. This framework is an initiative model based on the ECA rules, which can trigger kinds of activities initiatively under certain conditions according to the ECA rules. This meets the needs of our multi-Agent coalition system. In addition, for the safety of the system, a message encrypting technology is designed in the JMS operating module. The framework of an Agent's activities body is made up of three parts: the ECA rules obtainment module, the ECA rules instantiation module and the ECA rules executing module. At the last of the paper, the main classes for the modules in this framework and some user interfaces are given.