A BDI Model Based On Fuzzy Wavelet Neural Network

The BDI model which employs the modal logic to describe the agent's belief, desire, and intention reasons based on symbolic logic. It appears that this model has bad calculability, and it is unable to solve the fuzzy and non-linear problem. Using fuzzy neural network to realize the BDI model is possible for solving the problems above, but the fuzzy membership function of the fuzzy neural network can't be adjusted all the same, so it is limited to be used in such a non-linear BDI model for learning.The fuzzy wave neural network is used to realize the BDI model in this thesis. So the mental action of an agent can be realized by the neural calculation instead of complex symbolic logic. Therefore, using this model can overcome the defects that the symbolic logic has bad calculability and can't solve the non-linear problem. This model also improves the learning ability and non-linear processing ability of the neural network by adjusting the wavelet membership function.Firstly, the structure and learning algorithm of the BDI model which based on the fuzzy wavelet neural network are given. Then, the initialization and the general ability of the new model are discussed. Finally, the model is applied to solve the pursuit evasion problem. The results of simulation show that this model is feasible.