BY LAND, SEA, AIR, OR SPACE: CREATING COMPLEX SYSTEMS TO FUNCTION SEAMLESSLY
As vehicles and other cyber-physical systems become increasingly complex, there's a critical need for better methods of predicting the behavior of the physical processes and ensuring that the integrated computational components obey physical world constraints, such as control laws for flight surfaces or anti-lock brakes. That's where ISIS is playing a major role in advancing solutions.
Among the hardest problems facing researchers and engineers are those associated with producing robust and secure integration of physical and computational processes for cyber-physical systems, which deliver advanced capabilities in airplanes, cars, spacecraft, smartphones, and even smart power grids. Although their use is ubiquitous—with well over 90 percent of all microprocessors now used for cyber-physical systems it's historically been hard to validate and verify these systems due to the lack of theoretical foundations and automated tools for testing tightly integrated computational and physical systems.
To address the challenges of cyber-physical system validation and verification, Xenofon Koutsoukos, associate professor of computer science and computer engineering, and Janos Sztipanovits, ISIS director and E. Bronson Ingram Professor of Engineering, are leading the Science of Integration for Cyber-Physical Systems, funded by the National Science Foundation (NSF).
This project is creating model-integrated computing techniques, which enable the design, analysis, and integration of complex cyber-physical systems using automated tools. These tools will enable incremental validation and verification of key system properties, such as functional correctness, safety, and stability, so these systems need not be built and retested from scratch to accommodate every change.
This project is also focusing on how to combine disparate model-integrated computing tools into an open tool integration framework that cyber-physical system practitioners and engineers can apply to develop and sustain complex systems more rapidly and reliably throughout their life cycles, Koutsoukos said. These integrated tools are essential to aid in building and assuring future safety and mission-critical cyber-physical applications, such as autonomous air and ground vehicles.
Together with General Motors (GM), the University of Maryland, and the University of Notre Dame, Koutsoukos, Sztipanovits, and their ISIS team are building two platforms. The first is an experimentation platform consisting of a commercial simulator used by GM to design and test automotive functionality, such as adaptive cruise control and lane change warning. The network and the processors are real, but the car is simulated.
"This method allows us to address key implementation effects, such as network delays and processor failures, before designing the system," Koutsoukos says. "In the same way that operating system and application software in a computer or smartphone must be updated regularly, auto companies like GM must update car software. The moment that happens, we must ensure that everything continues to work without creating unforeseen problems." A key challenge is accommodating such updates while keeping the system safe, Koutsoukos added.
The second platform is a group of quadrotor helicopters that perform collaborative tasks, such as formation flying. More flexible than conventional drones, these helicopters have many potential applications, such as assisting in search-and-rescue operations, tracking smugglers, and controlling vehicle traffic.
Cyber-physical system innovations like these have huge implications for the way vehicles are made and how they move around. There are also many potential applications for devices in other cyber-physical domains, such as insulin pumps or pacemakers in the medical field.
ISIS has long been a leader in the field of model-integrated computing, pioneering domain-specific modeling languages, formal models and model transformations, and integrated tools to support simulation, analysis, and quality assurance, explained Sztipanovits. The modeling techniques and tools developed at ISIS as part of the NSF Science of Integration for Cyber-Physical Systems project helps predict and evaluate how different vehicle hardware and software components will interact.
"By building computer models, we are better able to understand the complexity and interaction of different cyber-physical components and to divide big problems into smaller problems so their solutions become more tractable and affordable," said Koutsoukos.
Understanding how to transform cyber-physical system integration from a "black art" into an engineering discipline based on sound theoretical foundations and scalable tools is the lofty goal of much ongoing cyber-physical system work at ISIS. Model-integrated computing is the future of complex cyber-physical systems, such as aircraft, spacecraft, and cars, Sztipanovits said.
When modern companies design an aircraft or a car, they buy computing devices and processors they plan to use for the lifetime of the product, all integrated after significant testing, which is expensive and time-consuming. "The moment anything is changed, all the testing is invalidated," he added. "That's the problem we're trying to address how not to do it from scratch over and over again."
Educational activities associated with the NSF-funded project are yielding new courses at Vanderbilt and new curricula for the cyber-physical systems field. "We must produce much new and revamped course material to educate the next generation of engineers needed to support rapid innovation," Koutsoukos said. At least one ISIS graduate student on the project team spent a summer working at GM.
Another goal of the project is to disseminate research results and foster an international community around cyber-physical systems. This ecosystem includes workshops, invited talks, plenary sessions, an advisory board, and increased collaboration amongst research groups around the world.
"The hope is that all of these technologies will be able to improve cyber-physical system safety and reliability with fewer recalls," Koutsoukos said. "Our model-integrated computing tools aim to minimize such problems when developing cyber-physical systems. It's the future and it's here to stay."