Verifiably Safe Autonomy for Cyber-Physical Systems
Abstract
This thesis demonstrates that autonomous cyber-physical systems that use machine learning for control are amenable to formal verification. Cyber-physical systems, such as autonomous vehicles and medical devices, are increasingly common and increasingly autonomous. Designing safe cyber-physical systems is difficult because of the interaction between the discrete dynamics of control software and the continuous dynamics of the vehicles physical movement. Designing safe autonomous cyber-physical systems is even more difficult because of the interaction between classical controls software and machine learning components.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 2018
- Accession Number
- AD1173982
Entities
People
- Nathan Fulton
Organizations
- Carnegie Mellon University