A novel guidance law with refining fuzzy membership functions and applying neural optimization algorithms alternatively

In this research, we proposed a guidance law by refining the fuzzy membership functions and applying neural optimization algorithms alternatively in each step, such as Gradient Descent (GD) and Levenberg-Marquardt (LM) methods. The boundaries of fuzzy error membership functions are lowered to increase the system performance. Besides, the error rate and output are divided into six types of models for trade-off. Both missile turning rate time constant and radome aberration error slope are taken into consideration. Comparisons with the proportion navigation (PN), fuzzy and the proposed methods are also made; we note the performances obtained by the proposed method are better. Besides, the acceleration commands are lower than the previous research and without oscillation or diverging effects at the final stage of interception. Thus even there are system parameter variation effects; the guidance loops can always track the target, and yields better guidance performances.