Intelligent Missile Guidance by Using Adaptive Recurrent Neural Networks

Abstract

In this paper, an adaptive recurrent neural network (RNN) controller is proposed for missile guidance. We address the problem of one agent (defending missiles) and one target (incoming missiles) in air battle scenario. The RNN controller is designed to force an agent (or defending missile) toward a target (or incoming missile), and a monitoring controller is also designed to reduce the error between the RNN controller and ideal one. The former is the main controller that can be easily designed. Its weighting factors are activated to dispatch the agent toward the target. By using the Lyapunov constraints, we update the weighting factors for the proposed RNN controller to guarantee the stability of the path evolution (or planning) system. Excellent simulation results are obtained by using this new approach for missile guidance, which show that our RNN has the lowest average miss distance (MD) among the several techniques.