New UTA research will automatically check for bugs in cyber-physical systems

The development environments in which cyber-physical systems are created must be correct if they are to work safely and reliably. Two computer scientists at The University of Texas at Arlington are using formal, methods-based testing to automatically discover bugs that will help improve these development environments.

Development environments allow engineers to use desktop software to design, model, simulate and analyze cyber-physical systems.

Taylor Johnson, an assistant professor in the Computer Science and Engineering Department, and co-PI Christoph Csallner, an associate professor in that department, will investigate how to automate improvement of development environments for cyber-physical systems with a $498,437 grant from the National Science Foundation.

Many software tools employ code generation methods to translate the model to the target hardware of whatever computer or microcontroller the software will run on in the actual system. This project aims to find defects in the development environments at several different stages, from the simulation stage through the code generation and compilation stages. This is accomplished through a technique known as randomized differential testing, where programs are randomly generated, translated with different tools - such as code generators and compilers, executed and then their outputs are compared.

"The results of all of the output should match. If not, there might be a bug somewhere in the system," Johnson said. "We must be able to randomly generate programs and cyber-physical systems' models, which we do using hybrid systems that have both continuous (how the physical world evolves) and discrete (model) behaviors."

This research will address several challenging theoretical and practical concerns novel to using this technique in cyber-physical systems, such as comparing simulation results to one another and approximating real numbers using floating-point numbers. Csallner will apply his expertise in software testing to the project, including prior work on randomly generating Java programs.

Khosrow Behbehani, dean of the College of Engineering, said Johnson and Csallner's work will make important contributions to the knowledge base of developing cyber-physical systems.

"Cyber-physical systems are prevalent in our lives, from automobiles, to aircraft, to many other systems. It is critical that we have the best possible development environments to ensure that they are reliable," Behbehani said. "Dr. Johnson and Dr. Csallner can make a significant contribution in giving future developers confidence to create with the knowledge that errors will not be introduced in the development phase."

Johnson is creating a niche in research related to cyber-physical systems. He currently has a $174,634 National Science Foundation grant, a $499,546 Air Force Research Laboratory grant and a $397,807 Air Force Office of Scientific Research grant all in this area of expertise.

Csallner was part of a recent $500,000 National Science Foundation grant that helped debug database-centric applications.

Provided by University of Texas at Arlington