Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

AVS 2018 Slides


Published on

Phil Koopman panel slides for AVS 2018. Briefly describes some approaches we take at Edge Case Research for self-driving car safety.

Published in: Technology
  • Be the first to comment

  • Be the first to like this

AVS 2018 Slides

  1. 1. Elements of Autonomous Vehicle Safety (The Super-Short Version) AVS 2018 © 2018 Edge Case Research LLC Prof. Philip Koopman @PhilKoopman
  2. 2. 2© 2017 Edge Case Research LLC The Real World: Heavy Tail Distribution Common Things Seen In Testing Edge Cases Not Seen In Testing (Heavy Tail Distribution)
  3. 3. 3© 2017 Edge Case Research LLC  Low integrity Doer  E.g., based on training data  High integrity Checker  Use traditional safety methods Planning and Control: Doer/Checker UNSAFE! Checker Shutdown Doer/Checker Pair Low SIL High SIL Simple Safety Envelope Checker “AI”
  4. 4. 4© 2017 Edge Case Research LLC Pseudo-realistic degradation Move and insert objects Perception: Robustness via Augmentation Pedestrian Missed: Gaussian Noise + Black Car Pedestrian Missed: Gaussian Blur Recorded scene Dark vehicle added to scene Synthesized scene
  5. 5. 5© 2017 Edge Case Research LLC Conventional V-based software safety  ISO 26262, SOTIF forsafety requirements Rapid safety case update tooling  STPA plus augmented fault trees and GSN Architectural safety patterns  Doer/Checker for a fail silent channel  Multi-channel approach for safing mission Robustness/stress testing  Traditional robustness testing  Object & event edge case “zoo” Other Techniques