DECENTRALIZED DYNAMIC PROGRAMMING FOR BALLISTIC MISSILE DEFENSE AND OTHER APPLICATIONS

THIS PROPOSAL CONSIDERS A NEW APPROACH TO THE \ DESIGN AND REAL TIME CONTROL OF DISTRIBUTED COMPUTER SYSTEMS FOR BALLISTIC MISSTLE DEFENSE. THIS APPROACH, CALLED DECENTRALIZED DYNAMIC PROGRAMMING, IS IDEALLY SUITED TO PROBLEMS IN WHICH INFORMATION PROCESSING, AND DECISION MAKING MUST BE CARRIED OUT AT SEVERAL INTERCONNECTED LOCATIONS, WHERE INFORMATION CAN BE TRANSMITTED BETWEEN THE COMPUTING RESOURCES AVAILABLE AT THE METHOD IS BASED ON NEW CONCEPTS OF DECENTRALIZED THEORY, AND IT IS APPLICABLE TO DYNAMIC,STOCHASTIC SITUATIONS AS WAS TO PURELY STATIC PROBLEMS. THE APPROACH IS BASED ON FORMULATING AN OVERALL OBJECTIVE FUNCTION, WHICH IS TO BE MAXIMIZED, THAT ACCOUNTS AT ALL LOCATIONS AS WELL AS THE COSTS FOR COMMUNTCATION OR TRANSFER OF RESOURCES OR WORK. THE OBJECTIVE IS THEN MAXIMIZED ACCOUNTING FOR THE REQUIREMENTS AND CAPABILITIES OF EACH LOCATION. A UNIQUE FEATURE OF THE METHOD IS THAT MAXIMIZINATION IS CARRIED OUT IN A DECENTRALIZED FASHION. EACH LOCATION IS ABLE TO SOLVE THE PORTION OF THE PROBLEM THAT INVOLVES IT. ONLY CERTAIN ESSENTIAL COORDINATION INFORMATION MUST BE TRANSMITTED TO OTHER LOCATIONS. THIS RESULTS IN AN EFFICIENT AND EFFECTIVE PROCEDURE.