Dynamic programming is a recursive computationaly efficient optimization technique for solving a class of multi-stage decision problems. It provides the optimum solution at a much smaller computational cost than exhaustive search. The dynamic programming algorithm (DPA) has potential applications to a number of important strategic defense problems including: multi-target tracking, data fusion, and target classification. In spite of its potential, the DPA rarely has been applied to real-time signal processing problems due to its computational requirements. Processing technology needs to be developed to realize practical applications of the DPA. The DPA lends itself to implementation using parallel processing architectures with local interconnections, such as systolic arrays. Extremely high processing rates can be achieved in principle by tapping the power of parallel computing. Very little work has been done to date to explore the optimum parallel implementation of the DPA. Research is being undertaken to develop a parallel ultra-high-speed DPA processor focussing on the detection and tracking of a large number of targets from a sequence of infrared images.
