Search

A neural-network-based synthesis of an optimal midcourse guidance law is presented in this study. We use a set of two neural networks; the first network called a “critic” outputs the Lagrange's multipliers arising in an optimal control formulation and second network, called an “action” network, outputs the optimal guidance/control. The system equations, the optimality conditions, the costate equations are used in conjunction with the network outputs to provide the targets for the neural networks. When the critic and action network are mutually consistent, the output of the action network yields optimal guidance/control. Numerical results for a number of scenarios show that the network performance is excellent. Corroboration for optimality is provided by comparisons of the numerical solutions using a shooting method for a number of scenarios.

We investigate the use of an 'adaptive critic' based controller to steer an agile missile with a constraint on the angle of attack from various initial Mach numbers to a given final Mach number in minimum time while completely reversing its flightpath angle. We use neural networks with a two-network structure called 'adaptive critic' to carry out the optimization process. This structure obtains an optimal controller through solving Hamiltonian equations. This approach needs no external training; each network along with the optimality equations generates the output for the other network. When the outputs are mutually consistent, the controller output is optimal. Though the networks are trained off-line, the resulting control is a feedback control.