ABDUCTIVE/EXPERT SYSTEM CONTROLLER

THE OBJECTIVE OF THIS EFFORT IS TO DEMONSTRATE A PROTOTYPE CONTROLLER FOR A HOWITZER USING A HYBRID NETWORK AND PRODUCTION RULE SYSTEM. THE RESULTING ADAPTIVE CONTROLLER WILL DEMONSTRATE THE FEASIBILITY OF DEVELOPING A REVOLUTIONARY CONTROLLER CAPABLE OF ADAPTING IN A DYNAMIC ENVIRONMENT, MONITORING AND DIAGNOSING BEHAVIOR, AND PREDICTING FUTURE FAULTS AND MAINTENANCE REQUIREMENTS. ABTECH HAS DEVELOPED AN INNOVATIVE NETWORK APPROACH, CALLED ABDUCTIVE NETWORKS, BASED ON MANY YEARS OF RESEARCH IN NEURAL NETWORKS AND ADVANCED STATISTICS. ABDUCTIVE NETWORKS ARE SIMPLY NETWORKS OF VERY POWERFUL NODE TYPES SUCH AS HIGH-ORDER POLYNOMIAL EQUATIONS. THIS PROPOSAL FIRST INTRODUCES AIM - A PRACTICAL SUPERVISED LEARNING TOOL FOR SYNTHESIZING ABDUCTIVE NETWORKS FROM DATABASES OF EXAMPLES. IT DISCUSSES HOW NETWORKS CAN BE APPLIED TO SUBSTANTIALLY INCREASE THE CONTROL AND DIAGNOSTIC CAPABILITIES OF SYSTEMS IN A PRACTICAL AND VERY COST-EFFECTIVE MANNER. IT THEN DETAILS A SPECIFIC PROJECT TO APPLY THE EXISTING AIM SOFTWARE AND A PRODUCTION RULE SYSTEM TO AN ARTILLERY SUBSYSTEM APPLICATION.