Abstract: For UCAV having the capability to deal with autonomous air combat, the flight dynamics model with the overload input and the 3-dimensional proportional navigation guidance model were established. Based on artificial neural network, an evasive maneuver decision was presented for avoiding incoming missiles. The degrees of freedom for UCAV-missile system were reduced by coordinate transformation, which simplified the complicated model as a non-linear model with relatively small amount of input and a single output. After the neural network samples were generated and trained, evasive results could be directly predicted from the relationship of positions between UCAV and missile through neural network, providing real-time evasive strategies to UCAV. Simulation example was also presented to verify the evasive method’s effectiveness.

Key words: UCAV, autonomous air combat, neural networks, maneuvers decision, evasive strategy