Framework for Autonomous Optimization

Phoenix Integration and MIT propose to create a novel autonomous optimization tool and application programming interface (API). The API will demonstrate the ability to link to many optimization algorithms, both open source and proprietary, as well as to framework tools that carry optimization algorithms within them. It will also allow users to connect their engineering models to it conveniently. The API will be available both as a cross-platform standalone product and as part of ModelCenter, an engineering integration and trade study environment. In addition to and included within the API will be techniques to perform optimization autonomously by providing a management layer which globally adjusts the run in an intelligent fashion. Thus, it will categorize problems to understand effective solution techniques for them, try many algorithms during a run, change the settings on single algorithms so they run more productively, adaptively learn which techniques worked and which didn't, and inquire of the user insight that may help the optimizer reach its destination sooner. A database of prior runs will be built to help facilitate these features. The management layer will also help the user understand errors that take place, log appropriately, and prevent failures.