A NEURAL NETWORK SOLUTION TO THE REAL-TIME OPTIMAL ALLOCATION OF MARINE CORPS TACTICAL AND C3I ASSETS

THE EFFICIENCY WITH WHICH ASSETS ARE ALLOCATED DURING COMBAT HAS USUALLY PLAYED A DOMINANT ROLE IN THE OUTCOME OF THE CONFLICT -OFTEN ECLIPSING ALL OTHER FACTORS, INCLUDING VARIANCES IN FORCE STRUCTURE. THE SOLUTION OF OPTIMAL ALLOCATION PROBLEMS IS NO LESS CRITICAL TO THE DESIGN AND OPERATION OF C3I NETWORKS. THUS, THE ABILITY TO PERFORM THIS ENABLING TASK AS EFFICIENTLY AS POSSIBLE IS VITAL. WE PROPOSE TO DEMONSTRATE THE FEASIBILITY OF USING A HYBRID ARTIFICIAL NEURAL NETWORK (ANN) ARCHITECTURE TO INTEGRATE TACTICAL INFORMATION SUCH AS FRIENDLY AND ENEMY FORCE STRENGTH AND DISPOSITIONS, AND THEN USE THAT INFORMATION TO CALCULATE THE OPTICAL ALLOCATIONS OF THE COMMAND'S COMBAT ASSETS SUBJECT TO A GIVEN SET OF CONSTRAINTS. OUR METHODOLOGY HAS BEEN SUCCESSFULLY DEMONSTRATED ON A VARIETY OF TACTICAL ALLOCATION PROBLEMS, INCLUDING COMMUNICATION RELAY SITE LAYOUT AND TERRAIN-SENSITIVE MOVEMENT PLANNING. THIS METHODOLOGY USES OUR PROPRIETARY COMBINATIONAL OPTIMIZATION ALGORITHMS. THE HEURISTICS EMPLOYED IN OUR NETWORK GENERATION ROUTINES RENDER THE RESULTING ANN MORE ROBUST WITH RESPECT TO LOCAL MINIMA AND INITIAL CONDITION SPECIFICATION THAN ANY OTHER DESIGN THAT HAS BEEN REPORTED. FURTHER, FOR LARGE SCALE PROBLEMS, OUR ALGORITHM IS MORE COMPUTATIONALLY EFFICIENT THAN TRADITIONAL ALTERNATIVES, EVEN ON SERIAL DIGITAL PROCESSORS.