This Small Grant for Exploratory Research (SGER) provides funding for a new but untested framework for closed-loop planning and control systems, focusing on dynamic information gathering, optimization and control in a distributed environment. The new framework would transform the paradigm of open-loop planning to a new paradigm of closed-loop planning and would create distributed, extensible systems that can be supported by high throughput computing and are responsive in real time to dynamically changing environments in which there is uncertainty and incomplete information. The nested partitions optimization method will be studied for application in this framework because it combines adaptive global sampling with local heuristic searches, has multiple tasks that can be distributed, and is flexible in the degree of parallelism. A key aspect to be explored will be dynamic evolution of neighborhoods for decoupling in distributed optimization and the symbiotic evolution of sensors that measure dynamically changing coupling and uncertainty in a distributed environment. Measures of completeness of knowledge coupling and consistency of information sharing will be developed that promote accommodation of uncertainty in distributed optimization.
This work will address the growing need for quick response, extensibility and adaptability in planning applications and the dynamic linking of planning with control, while exploring the integration of the theory and methods of production systems engineering with those of control systems engineering. The research will be focused on production planning, but the results are expected to be important in dynamic planning in other areas such as health care and transportation, resulting in more efficient use of resources. The work will also contribute to advances in distributed simulation, distributed optimization, distributed control and their integration.
