Real-Time Optimization in Complex Stochastic Environment

The research reported here aims at enabling a systematic on-line use of optimization techniques for real-time applications in complex stochastic environments that recognizes requirements for new generations of systems critical to the national infrastructure and consistent with the emerging information-based, network-centric view of warfare. The main outcomes of the project are a Asynchronous, event-driven distributed optimization algorithms with the ability to escape frequently occurring local equilibria. b Novel on-line trajectory optimization schemes for cooperative multi-agent systems which are scalable and robust with respect to the uncertainty model used. c A general-purpose event-driven receding-horizon optimization framework. d A general unified Infinitesimal Perturbation Analysis framework for efficient gradient-based optimization of complex stochastic systems.