Accession Number : ADP006335
Title : Neural Network Solutions to Mathematical Models of Parallel Search for Optimal Trajectory Generation,
Corporate Author : SMITHS INDUSTRIES AEROSPACE AND DEFENSE SYSTEMS INC GRAND RAPIDS MI GRAND RAP IDS DIV
Personal Author(s) : Reibling, Lyle A.
Report Date : SEP 1991
Pagination or Media Count : 9
Abstract : This paper describes the problem and solution method in envisioning massively parallel architectures; which incorporate mathematical physics models for trajectory generation applications. A difficult problem in search applications is computing the optimal aircraft trajectory in real time onboard a high performance aircraft, where the objective is to increase th aircraft survivability and mission effectiveness by penetrating enemy threats and minimizing threat radar exposure. Optimal trajectory generation is achieved by searching all possible paths in a multidimensional search space for the path with the smallest accumulated performance measure, which represents total threat exposure. This search problem has many state variables in a large space which places severe demands on computational resources, requiring real-time solutions unavailable from current technology. The mathematical model which is the basis for this architecture is based on an application of electrostatic field theory.
Descriptors : ACCUMULATION, AIRCRAFT, COMPUTATIONS, COMPUTER ARCHITECTURE, ELECTROSTATIC FIELDS, ENEMY, EXPOSURE(GENERAL), FIELD THEORY, HIGH RATE, MATHEMATICAL MODELS, MEASUREMENT, MISSIONS, NEURAL NETS, OPTIMIZATION, PARALLEL ORIENTATION, PARALLEL PROCESSING, PERFORMANCE(ENGINEERING), PHYSICS, RADAR, REAL TIME, RESOURCES, SEARCHING, SOLUTIONS(GENERAL), SURVIVABILITY, THREATS, TRAJECTORIES, VARIABLES.
Subject Categories : Cybernetics
Distribution Statement : APPROVED FOR PUBLIC RELEASE