Can We Really Use Machine Learning in Safety Critical Systems?

Richard Murray
California Institute of Technology
Control and Dynamical Systems

Machine learning has the promise of revolutionizing our approach to many problems in which computers have typically not performed well (such as image classification, speech recognition, and related problems). It is being extended and applied to systems in which these problems are part of a much larger set of complex decision-making tasks, often in safety- and mission-critical systems. In this talk I will discuss the methods used in engineering applications to provide high confidence in aerospace and other systems, and the challenges of adopting those techniques to include "machine learned” components.

Presentation (PDF File)

Back to Intersections between Control, Learning and Optimization