Study On Behavior Of Agents In Uncertain Environment

Study on the behavior of Agent in the uncertain environment is an important problem in Artificial Intelligence. With the development of computer science and technology, the operating environment of computer systems is more and more complex, which requires the higher intelligence of computer system. Therefore, the multi-agent system gradually comes into people’s sight. The multi-agent system is applied more and more widely, and the agent needs to work in a complex environment, especially, in the uncertain environment. The traditional agent in the uncertain environment is difficult to have the good and suitable behaviors and it lacks the ability to represent the uncertain knowledge, which leads agent hard to estimate its tasks and accomplish the desired plans. It is urgently needed to study the behavior of agent in an uncertain environment. The research of agent technology mainly is divided into the study of a single agent and the interactions between agents.This paper mainly studies the single agent and puts forward a kind of improved BDI Model, which integrates with Bayesian Network. The improved model changes the knowledge structure of traditional BDI Model. It reasons as Bayesian Network does, as well as plans.This improved model makes full use of the BDI model and Bayesian Network, so that the agent in an uncertain environment can express the cognition of uncertainty environment, as well as estimate the desires and complete decisions.This paper uses Java language in Eclipse platform to simulate the cultivation of crop. The experiment uses a number of sets of data to simulate expected measures being taken when the crop is facing various disasters. The result of simulation verifies that the model can work in an uncertain environment and can make up the deficiency of agent.BDI model integrated with Bayesian Network in the static uncertainty environment can play a good role.When the environment is dynamic, the effect is relatively poor. This can be solved by introducing dynamic Bayesian Network in the BDI model.