OPERATIONS PLANNING WITH NEURAL NETWORKS

ADVANCEMENTS IN OPERATIONS PLANNING ALGORITHMS ARE SOUGHT FOR USE IN SIMULATIONS WITH THE JPL TIME WARP OPERATING SYSTEM (TWOS) EXECUTING ON PARALLEL COMPUTERS. OUR OBJECTIVE IS TO DETERMINE THE FEASIBILITY OF OPERATIONS PLANNING WITH NEURAL NETWORKS AS THE DECISION MAKING MECHANISM. THE EFFORT WILL CONSIST OF CONSTRUCTION OF A NETWORK MODEL, MODEL INITIALIZATION, AND TESTING. EXPERIMENTS INVOLVING BIDIRECTIONAL ASSOCIATIVE MEMORIES (BAM) AS THE NEURAL NETWORK WILL BE TESTED IN BOTH SEQUENTIAL AND PARALLEL DISCRETE EVENT SIMULATION (PDES) ENVIRONMENTS. RESULTS OF ACCURATE DECISION MAKING AND INCREASED SIMULATION SPEED COULD LEAD TO FUTURE DEVELOPMENT OF A PDES PRODUCTION VERSION AS A PLANNING TOOL. GENERAL APPLICATION TO OPERATIONS PLANNING CAN PROVIDE NASA WITH TIME AND COST SAVINGS BY BETTER TASK SCHEDULING, COMMUNICATION, FACILITY SCHEDULING, MANPOWER USAGE AND RETAINING "CORPORATE MEMORY". SUCH CAPABILITY CAN BE EXTENDED COMMERCIALLY TO PLANNING FOR TRANSPORTATION INDUSTRIES, LOGISTICS, PRODUCTION FACTORIES AND CONSTRUCTION.