Cyber-physical systems (CPSs) are physical systems controlled by embedded software. Examples include airplanes, ground vehicles, unmanned air vehicles, ships, and submarines. The functions that CPS software controls are critical for both safety and timing: producing the wrong results or even the right results at the wrong time can have disastrous consequences. The challenge is that traditional tools and techniques that have been engaged to ensure the timeliness and correctness of software are costly and ineffective when applied to CPSs. Fortunately, the SEI together with other departments of CMU have a long history of deep expertise in real-time systems, formal techniques (such as model checking) for ensuring logical correctness of software, robotics, ultra-large-scale (ULS) systems, and big-data analytics.
Cyber-physical systems are one type of ULS system. ULS systems are systems of ultra-large scale along many dimensions, including interdependent webs of software-intensive systems, people, policies, cultures, and economics. They are characterized by decentralization; inherently conflicting, unpredictable, and diverse requirements; continuous evolution and deployment; heterogeneous, inconsistent, and changing elements; erosion of the people/system boundary; and routine failures. Despite such challenging characteristics, ULS systems must reliably demonstrate desirable behavior.
The Cyber-Physical and ULS Systems initiative develops principles and technology to understand, control, and bound the behavior of cyber-physical-social systems. We focus on two areas: high-confidence CPSs and socio-adaptive systems.
One research focus is high-confidence cyber-physical systems. The goal is to enable efficient development of autonomous CPSs whose distributed elements operate in a provably correct and timely manner and consequently whose collective behavior can be predicted and relied on. This entails demonstrating scalable algorithms for functional analysis of real-time software, techniques for controlling effects of multicore memory access on CPS real-time behavior, and techniques for assuring distributed autonomous coordination.
A second focus is socio-adaptive systems. The objective is to work with the human element in systems to enable optimal responses to changes in resource capacity. Using computational mechanism design to elicit information about changing needs and resources, a socio-adaptive system computes optimal allocation of resources and optimizes a decentralized quality of service.