Re-entry guidance by threshold network storage of precomputed optimum commands.

This paper describes a technique for the onboard generation of re-entry guidance commands based on the prediction of landing footprints. This procedure does not require the large amounts of very rapid real-time computation normally needed by such predictive guidance schemes. The re-entry guidance problem is treated as a surface approximation problem where the collection of points relating optimum guidance commands to the N state variables forms a surface in (N +1) dimensional space. Each of these points is determined before a particular mission by computations involving the integration of appropriate equations of motion. Threshold networks are designed to provide a piece wise linear surface which represents an approximate solution (with arbitrarily small errors) to the complex guidance equations. One such surface is stored for each control variable, and the resulting threshold networks can be used for onboard guidance. Several simulations of re-entry trajectories are described, with the threshold network furnishing guidance commands. The range covered is several thousand miles and, in all cases, landing accuracies are a few hundred feet, which is of the same order as the errors in the computer generated data.