Project Title:

A Neural-Net Approach to Space Vehicle Guidance

09.02-3474
NAS1-19004
A Neural-Net Approach to Space Vehicle Guidance
Charles River Analytics, Inc.
        55 Wheeler Street
        Cambridge, MA 02138
                Alper K. Caglayan  (617-491-3474)

Abstract:

        The numerical algorithms involved in the solution of optimum trajectory and
guidance problems are too complex for on-line application with advanced space vehicles.
Hence, the current approach to the development of real-time guidance is to use approximation
theory to obtain closed-loop guidance laws. Neural networks offer an alternative
to the derivation and implementation of guidance laws. This project will formulate
the space vehicle guidance problem using a neural network approach and find the appropriate
neural net architecture for modelling optimum guidance trajectories. It will train
the developed network with a database of optimum guidance trajectories and demonstrate
its performance as an on-line classifier. Such a neural-network-based guidance approach
can readily adapt to environmental uncertainties such as those encountered by an
AOTV during atmospheric maneuvers.

Potential Commercial Application:

Potential Commercial Applications: The commercial application would be a front end
for neural-network software packages and computers for incorporating a priority-system
knowledge base into the selection of processing elements and interconnect structures.