VeHICaL — or Verified Human Interfaces, Control and Learning for Semi-Autonomous Systems — is an interdisciplinary project involving UC Berkeley researchers that seeks to develop improved ways to build human cyber-physical systems, like self-driving cars and drones.
This project has received a five-year $4.6 million grant from the US National Science Foundation to conduct its research and relies on cross-campus collaboration with eight professors from UC Berkeley, Caltech and University of North Carolina at Chapel Hill.
“Cyber-physical systems really have the potential to improve a lot of things about everyday life,” said Sanjit Seshia, principal investigator of VeHICaL and UC Berkeley professor. “For instance, if we had reliable, dependable autonomous control (of) our cars, then potentially we could deal with problems like distracted driving.”
VeHICaL conducts both conceptual and experimental research, Seshia said. The conceptual component creates mathematical models of these complex systems and provable algorithms of semi-autonomous vehicles, according to Seshia, whereas the experimental aspect tests the algorithms in driving simulators and with drones in lab facilities.
So far, semi-autonomous systems have confronted difficulties regarding instances in which the machine cannot operate on its own and needs human input. The project hopes to confront the challenge by developing designs that will help determine when a driver should switch between manual and automated driving.
“We want to figure out how to design and build such systems so that they operate more safely than either human-only or machine-only systems,” said Caltech professor Richard Murray, co-principal investigator, in an email. “This means understanding what information the machine should provide to the human and vice versa.”
Additionally, the project is pursuing educational outreach at the graduate, undergraduate and pre-college level. An associated summer program is already ongoing, according to Seshia, while graduate students and postdoctoral researchers will later be recruited for VeHICaL.
Elsewhere, researchers are also working on various aspects of this topic, Seshia said, adding that he believes that the VeHICaL team is unique in that it has a diverse blend of expertise in fields such as robotics, privacy and security, control theory and cognitive science.
According to Gurdip Singh, NSF program director managing VeHICaL, the project was chosen by the NSF because it approached challenging problems and drew expertise from many different disciplines with a broad goal in mind.
“At Berkeley, we always try to make sure that our research projects have a broader impact, so it’s not just purely about research, but with an impact on education and society in general,” Seshia said.
Seshia added that VeHICaL is also unofficially collaborating with automotive companies and regulatory agencies — such as Toyota and the National Highway Traffic Safety Administration — to gain a perspective on the struggles manufacturers face in the industry.
Contact Fionce Siow and Charlene Jin at [email protected].