Verification Tools for Embedded Systems

We propose a five-year research initiative at Carnegie Mellon University to create a new generation of formal verification tools that can be integrated into design environments for the complex, high-assurance embedded systems that will be required by future military systems. These systems are increasingly distributed, complex dynamic systems that must operate with a high degree of autonomy and survivability in diverse and unpredictable environments. We will focus on the development of new verification methods and tools to provide a rigorous means for checking the integrity and correctness of designs for these systems before they are deployed on target platforms. Our initiative has three broad research thrusts: Verifying System Integrity. System integrity refers to the correctness of the embedded system operation with respect to the interactions among the distributed software and hardware components. The embedded system must satisfy synchronization constraints, resource constraints, and real-time constraints imposed by the implementation architecture and application requirements. Methods for performing formal verification of these types of constraints that have been successful in hardware and protocol applications will be extended to embedded system applications. Modeling the Environment. Embedded software systems for military applications interact in complex ways with physical systems and adverse environments. It is thus essential to capture correctly and effectively the continuous dynamics, feedback loops, and unpredictable features of the environment in the models used for formal verification. Our research will draw on recent developments in (i) hybrid system verification to integrate continuous state dynamics with discrete-state models used in formal verification; and (ii) probabilistic verification to model and evaluate critical system attributes such as survivability and fault tolerance. Usability. In addition to performing research on fundamental methods in formal verification, we will address the formidable barriers that have kept formal verification techniques beyond the reach of practicing design engineers. We propose to develop tools that will extract automatically models used for verification from models already being constructed by design engineers for simulation studies and software development. We will also develop new methods for presenting and interpreting the results from formal verification tools so that they can be used effectively to evaluate the system and to correct the sources of faulty behaviors. To assure this research initiative addresses the needs of real-time embedded systems in military applications, we have established a partnership with the Honeywell Technology Center (HTC), which will participate with in-kind matching funds. We will meet regularly with HTC researchers who will provide requirements specifications and review research plans and results. HTC will offer challenge problems from real avionics and combat systems and will use tools developed in this project on internal HTC projects to help with the evaluation and guidance of the CMU research. This initiative will also be an integral technical component of the newly formed High Dependability Computing Consortium, a partnership among Carnegie Mellon University, NASA, the Silicon Valley corporate sector, and other private and public sector organizations.