5/27/2020 Lizzie Roehrs, CSL
Written by Lizzie Roehrs, CSL
Distribution A. Approved for Public Release. Case Number: 88ABW-2020-1868
When flying roughly 35,000 feet above the surface of the Earth, airliner pilots rely on numerous control systems to ensure safety conditions such as the ones that prevent the aircraft from stalling. According to CSL Professor Sayan Mitra, the current approaches to testing this control software leave out important gaps. Mitra is working with industrial collaborators to develop an algorithm to test and certify software for certain aircraft control systems.“Currently, certification of complex control systems is based on standardized processes,” says Mitra, a professor of electrical and computer engineering. “It is useful, but the process itself does not guarantee that the end-product is safe or meets the desired requirements.”
This will be the first method that can evaluate gray-box systems. These are systems in which some of the structural information is known and can be exploited for analysis, while other information is not available and can only be approximated. For example, the adaptive cruise control system on a car contains both known and unknown information. The cruise control system knows to brake if it’s getting too close to the car ahead, but the actual rate of deceleration upon braking is unknown. This information is unknown because there are many outside factors that contribute to that rate. Such knowable and unknowable operations also exist in aircraft.
Mitra’s research will remove some of the “guesswork” from the analysis of control systems. Grey box system verification aims to combine theoretical structure with data to complete the analysis task. According to Mitra, a side effect of this type of analysis could be the determination of a set of conditions or assumptions under which the system is supposed to work correctly. Mitra’s algorithm aims to mathematically prove that a system can and does work the way it is designed to.
“Here, the dream is to develop algorithms or compilers that can take the code – say, the code for controlling an aircraft – and mathematically prove that the software does the right thing,” says Mitra. “Our team is developing tools – based on our algorithms – that show how this dream could become a reality for realistic vehicle control systems.”
The applications for the work could go far beyond just ground, sea, and air vehicles. According to Mitra, the world of software enabled control systems is huge and still growing.
“Effective technology for verification and certification could change how everything from autonomous cars to spacecraft to manufacturing systems are designed and tested,” says Mitra. “Usable and scalable verification technologies could broaden the population involved in building systems and applications by providing the languages, compilers, and tools for safe product development and experimentation.”
Distribution A. Approved for Public Release. Case Number: 88ABW-2020-1868