CPS V&V I&F Workshop 2019: Program

The CPS V&V Industrial Challenges and Foundations workshop includes a discussion session on safety challenges and hazards in CPS operations, as well as several talks from experts in industry and academia. These talks are meant to set the stage for discussions to be held at the workshop.

Note: this schedule is tentative and is subject to change. Check back here for future updates to the program and schedule.

Tuesday, Dec 10 (optional)

Carnegie Mellon University
13:00 Undergraduate students in the CMU Logical Foundations of Cyber-Physical Systems course will present their final projects to a panel of experts in CPS for the CPS V&V Grand Prix.

Wednesday, Dec 11

Gates-Hillman Center (GHC), Room 6115, Carnegie Mellon University
09:00 Welcome and Workshop Overview - André Platzer
09:15

Session I: Safety Assurance Challenges (Chair: Andre Platzer)

Constance Heitmeyer: Assuring the Safety and Correctness of Autonomous Systems - Challenges and Promising New Methods

Kristin Y. Rozier: Safe Learning: A Challenge Talk

10:15 Coffee Break
10:30

Session II: Perception Challenges (Chair: Georgios Fainekos)

Taylor Johnson: Challenges for Perception Verification in Autonomy

Nathan Fulton: Safe Learning with Visual Inputs

Lujo Bauer: On Evasion Attacks against Machine Learning in Practical Settings

12:00 Discussion I: Safety Challenges and Hazards in CPS Operations (Chair: Alwyn Goodloe)
12:30 Lunch Break
13:30

Session III: ML, Testing, and Falsification (Chair: Kristin Rozier)

Jim Kapinski: Application of light-weight verification methods to delivery robots

Galen Edward Mullins: Machine Learning for Testing and Evaluating Autonomous Systems

Georgios Fainekos: Gray-Box Adversarial Testing for Control Systems with Machine Learning Components

15:00 Coffee Break Discussion
15:30

Session IV: Verification Tools (Chair: Stefan Mitsch)

Bjorn Andersson: Mixed-Trust Computing - Guarding Untrusted Components in Learning Systems

Souradeep Dutta: Sherlock - A Tool for Verification of Neural Network Feedback Systems

Yasser Shoukry: Formal Verification of End-to-End Deep Reinforcement Learning

Hussein Sibai: Accelerating formal verification of autonomous multi-agent systems using symmetry transformations

17:30 Discussion II: Verification Challenges Summary (Chair: Stefan Mitsch)
18:00 Workshop end

Talks

The primary purpose of each talk should be to give context to and provide background for the full group discussions. In order to meet the goals of this workshop, we will feature two kinds of talks: solutions talks and challenge talks. These talks should generally adhere to the following guidelines.

Solution talks should introduce:

  1. The technical core idea behind the V&V and/or correct implementation approach.
  2. A qualitative critical assessment of its applications.
  3. Future challenges for safe implementation of CPS that can be connected with emerging techniques such that our community can focus on the most pressing bottlenecks with the biggest potential impacts.
Application challenge talks should introduce:
  1. Background on the fundamental technical characteristics of the particular application domain.
  2. An overview of current use of V&V and safe implementation techniques in the application domain.
  3. Future possibilities for V&V in the application domain, including a discussion on the blocking factors that need to be resolved.