Sign In

Communications of the ACM

ACM TechNews

Engineers Are on a Failure-Finding Mission

View as: Print Mobile App Share:
The team has shown the approach can root out failures in a variety of simulated autonomous systems.

MIT engineers have developed a new approach to quickly identifying a range of potential failures in autonomous systems.

Credit: iStock

An algorithm developed by Massachusetts Institute of Technology (MIT) researchers aims to identify potential failures in simulated autonomous systems prior to their real-world deployment.

The automated sampling algorithm also can determine how to fix the failures and suggest repairs.

Unlike other automated searches intended to detect the most severe system failures, the new algorithm can catch more subtle vulnerabilities as well.

The algorithm also can uncover hidden correlations in the system.

Additionally, researchers can go back through the chain of changes that resulted in a certain failure to determine how to fix it.

Said MIT's Chuchu Fan, "If you can predict a failure, you should be able to predict what to do to avoid that failure. Our method is now closing that loop."

From "Engineers Are on a Failure-Finding Mission"

MIT News (11/09/23) Jennifer Chu
View Full Article


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account