In the conditions of geopolitical instability, the problem of ensuring sustainability and dynamic development of the national innovation system is of critical importance. The article proposes a scientific and methodological approach to dynamic identification and forecasting of trajectories of technological innovation dissemination based on the synthesis of an intelligent decision support system. A formalized mathematical model of the national innovation system is used as a methodological basis, represented by a system of stochastic differential equations describing the interaction of knowledge flows, the intensity of technological search and the level of innovation diffusion. To control the model, a robust multi-loop cascade structure with a dynamic corrector and an internal model has been developed, the parameters of which are adjusted using the minimax optimality criterion and machine learning methods. Simulation modeling confirmed the effectiveness of the architecture, demonstrating a significant improvement in direct indicators of the quality of transient processes (control time and dynamic deviation) under conditions of changing parameters of the control object and external disturbances. The results of the study allow us to substantiate the effectiveness of public-private partnership measures as a key mechanism for mobilizing investments and to minimize dysfunctions in the development of innovative activities of actors in interdisciplinary scientific and technical programs.