Friday, November 10, 2023 | 11:00 AM
Uncas A. Whitaker Hall, 218
6760 Forest Park Pkwy, St. Louis, MO 63105, USA
Department of Electrical and Computer Engineering
University of Illinois at Urbana Champaign
Formal verification of deep learning models remains challenging, and yet they are integral in safety-critical autonomous systems. We present framework for certifying end-to-end safety of learning-enabled autonomous systems using perception contracts. The method flows from the observation that the role of learning-based perception is often state estimation, and that the error characteristics of such estimators can be succinct, testable, and amenable to formal analysis. Our framework constructs approximations of perception models, using system-level safety requirements and program analysis. Mathematically proving that a given perception model conforms to a contract will be a challenge, but empirical measures of conformance can provide confidence levels to safety claims. We will discuss recent applications of this framework in creating low-dimensional, intelligible contracts and end-to-end safety certificates for vision-based lane keeping, automated landing, and in runtime monitoring.