Article By Emily Cerf

With the support of a nearly $6 million grant from the National Science Foundation (NSF) through their Cyber-Physical Systems program, researchers at UC Santa Cruz will lead a five year, multi-institutional project to explore a new vision of engineering cyber-physical systems (CPSs). 

CPSs are highly complex systems that involve algorithms, networks, and physical components. Examples of CPSs include smart power grids, implantable medical devices, and transportation such as self-driving cars, the latter being the focus of this project. 

Project Goals

This project aims to rethink the modeling, analyzing, and designing of a new generation of intelligent transportation systems so the algorithms running them are adapted to computational constraints and the systems can run efficiently and reliably. The researchers will collaborate with industry and academic partners to advance CPSs both in research and education through strong training programs for high school and undergraduate students, with a particular focus on creating research opportunities for students from underrepresented backgrounds.

“​​This research will have direct impact in the rapidly growing, multi-billion dollar autonomous systems market,” said Ricardo Sanfelice, lead principal investigator on the project, professor of electrical and computer engineering, and director of the UCSC Baskin School of Engineering’s Cyber-Physical Systems Research Center (CPSRC). “We envision that our results will have a broad impact by improving the safety and reliability of transportation systems, such as aviation systems and self-driving vehicles, in particular, by reducing the carbon footprint of these systems, and training the workforce of the future in key CPSs science.”

Designing for adaptability 

CPSs face major engineering challenges from the computational limitations of traditional processors as well as the scale and diversity of physical components, which can be human-made structures and/or the natural landscape. In traditional systems, the computers are updated with information from the physical systems only periodically, meaning the system at certain points runs on old information which could jeopardize its safety and performance.

To solve these problems, this project will focus on codesigning the algorithms and hardware of CPSs so that the physics, hardware, and software are unified. The researchers will use results from verification, implementation, and testing of their new systems to redesign their algorithms, a process which will also happen in an automated fashion as the systems are running. This is unlike the current state-of-the-art systems, in which algorithms and hardware are typically not jointly designed, leading to a lengthy and costly verification process. 

This new model of providing feedback to CPSs will allow researchers to create systems that are much more adaptive than the current state of the art. The new control algorithms will adapt to the specification and the environment they are deployed on, learning and adjusting to key factors such as power consumption and execution time. The new hardware will be tailored to best provide feedback that can be used by these algorithms. 

“To meet the stringent requirements that intelligent transportation applications demand, such as performance and safety, the algorithms implemented in the control stack require advanced algorithms that exploit data to learn the environment, the physics, and the cyber,” Sanfelice said. “We believe this is the best way to enable CPSs to make close-to-optimal decisions.”

These new tools will reduce overall development cost and time for CPSs. The more accurate models will eliminate overprovision of hardware, and the new software will be open-source to enable broader reuse.