Science of Cyber-Physical System Integration

The objective of this project is to develop a science of integration for cyber-physical systems (CPS). The proposed research program has three focus areas: (1) foundations, (2) tools and tool architectures, (3) systems/experimental research. The project has pushed along several frontiers towards these overall objectives. In the following, we describe selected accomplishments:

Passivity-based design of automotive CPS
Developed and demonstrated passivity-based design method for control system design and integration. The framework has been studied and evaluated for both traditional control such as composition of throttle and brake control in Adaptive Cruise Control, and advanced control such as Semi-autonomous Driving. Evaluation is performed in an integrated end-to-end modeling and hardware-in-the-loop automotive simulation test bed developed by the research team. A transition plan has been created by GM R&D to evaluate the framework in GM product design.

Semantic Backplane for CPS Model Integration Platform
Introduced a horizontal Model Integration Platform for design of CPS which complements existing vertically integrated tool suites. The key innovation is a dynamic model integration language used for modeling cross-domain interactions using semantic interfaces to integrated domains. The Semantic Backplane is the formal framework for the Model Integration Platform. It includes the formal specification of semantic interfaces for design domains, the formal specification of composition semantics, and model transformations used in the model-based design flows. The formal framework is based on Formula, a Microsoft Research tool that was originated at Vanderbilt in a previous NSF funded project. The initial design of the Semantic Backplane was transitioned to a major DARPA CPS design automation project, Adaptive Vehicle Make (AVM).

Energy-Based Concepts of Passivity and Dissipativity in the Design of CPS
A major theoretical accomplishment for this project has been the development of tools of passivity and dissipitavity as a means for composable design in large scale cyber-physical systems. Passivity and dissipitavity quantify the property of a system to dissipate some of the energy that is either being supplied to it through external input or that is stored in the system. Classically developed for circuit theory, innovations to enable the application to CPS include passivity analysis of systems that are not accurately known, switched, hybrid and discrete-event systems, analysis in the presence of communication channels, and experimental determination of passivity.

Model-Based Systems Engineering for CPS
Developed a framework for cross-domain integrated modeling hubs for CPS and demonstrated applications to power grids, micro-robotics, energy efficient buildings, wireless sensor networks and vehicle management systems for next generation aircraft. A powerful tradeoff analysis tool (CONSOLOPTCAD) is integrated with CPS modeling integration hubs to enable broad exploration of the design space. In addition, an Interactive Tree Decomposition Tool is developed for reducing and managing system analysis complexity throughout the CPS design and operation cycle.

Education and Training
Through collaboration, academic members of the research team improved their understanding of industry design practices and real world constraints. GM hosted multiple visits of the research team, and sponsored 8 summer interns 2011 - 2014. Meanwhile, the research team brought advanced concepts to GM vehicle design practices, some of which are under active exploration to make them ready to transfer to engineering practices.