Domain-Specific Modeling

Abstract:

The objective of this research is an injection of new modeling techniques into the area of Cyber-Physical Systems (CPSs). The approach is to design new architectures for domain- specific modeling tools in order to permit feedback from analysis, validation, and verification engines to influence how CPSs are designed. This project outlines new research into the integration of existing, heterogeneous modeling languages in order to address problems in CPS design, rather than a single language used to design any CPS. Since many tools for analysis, validation, and verification focus on at most two of the three major components of CPS (communication, computation, and control), new paradigms in modeling are used to integrate tools early in the design process. The algorithms and software developed in this project run validation and verification tools on models, and then close the loop by using the tool outputs to automatically modify the system models. The satisfaction of design requirements in CPSs is of tremendous importance for tomorrow's societal technologies such as smart buildings, home healthcare, and water management. Perhaps the most compelling design requirements are those of safety, and CPSs such as autonomous vehicles exemplify this well. By involving a full-sized autonomous vehicle in this project, the validation and verification of safety requirements is tied to a concrete platform that is broadly understood. By involving students in the design of behaviors of the vehicle, the project exposes scientists and engineers of tomorrow to societal-scale problems, and tools to address them. The full-sized vehicle is involved in visits to local high schools in the Tucson area, where day-long activities permit students to develop their own decision algorithms to drive the car. Using the domain-specific tools from the project, selected designs are executed on the vehicle after they are verified for safety. Undergraduate and graduate courses are involved with the project in order to develop new modules for the autonomous car, which can be verified using the tools, and reused by high-school students as part of the project outreach. This year's focus is on new model transformation theory to synthesize the model transforms from failures in the verification code. Results to date include synthesized tests for model-based message-passing frameworks, which were used in an NSF REU Site "CatVehicle" for which the PI for this award is also PI.

Related Publications:

1. P. Morley, Warren, A., Rabb, E., Bunting, M., Whitsitt, S., and Sprinkle, J., "Generating a ROS/JAUS Bridge for an Autonomous Ground Vehicle", in The 13th Workshop on Domain-Specific Modeling, Indianapolis, IN, (in press).
2. S. Whitsitt and Sprinkle, J., "Modeling Autonomous Systems", AIAA Journal of Aerospace Information Systems, vol. 10, no. 8, p. 413, 2013.