In a large class of intelligent machines, one of the tasks of the knowledge‐based system components is to configure real‐time signal processing systems. These systems implement low‐level sensory or control algorithms according to the model of the system to be observed or controlled. If a change in the state of the model is detected, or the goal of the operation is modified, the real‐time signal processing schemes have to be dynamically restructured. This article describes a system which has been developed for the knowledge‐based generation and restructuring of real‐time signal processing systems in parallel computing environments. the system is based on the integration of a hierarchical representation technique, the corresponding interpretation method, and an execution environment, which includes a graph model of computations.