Invariant imbedding and parallelism in dynamic-programming for feedback-control

This paper investigates related areas such as invariant imbedding, state feedback, and numerical and parallel methods in order to specify the range of control problems amenable to a dynamic programming approach. Several forms of functional equations are classified according to different applications of the invariant imbedding principle and corresponding closed-loop control structures. Computational methods to implement these algorithms are described, and a complexity analysis is made to determine their effectiveness and to explain their application domain. The design of parallel algorithms is also considered. Alternative descriptions are compared; the frame of a distributed computational method delivering tabular feedback solutions is highlighted.