Главная
Study mode:
on
1
Intro
2
DARPA - Assured Autonomy Program Goal
3
Illustrating the challenge
4
Safety Assurance for Systems - State of Practice
5
VU ALC Project vision
6
Project activities
7
AUV as a CPS with LECS Learning-Erabled Components
8
LEC Verification: Reachability Analysis of Feedforward/Convolutional Neural Networks
9
Closed-Loop CPS with LECS: Verification Flow and Tools
10
UUV Closed-Loop Verification Results
11
ACAS-Xu Closed-Loop Scenario
12
ACAS-Xu Closed-Loop Verification Results
13
Perception Robustness Verification Target application Perception LE Component
14
Semantic Segmentation Robustness
15
Assurance Monitoring
16
Inductive Conformal Prediction (ICP)
17
Anomaly Detection
18
Exchangeability Martingales
19
Distribution Shift Detection Example
20
Distribution Shift Detection in adversarial scenarios
21
Tool architecture coverage
22
ALC Toolchain Our approach
23
ALC Design Workflow Specialized for LEC development
24
Modeling Blocks, Systems, Training, & Execution
25
System Modeling
26
System architecture SysML block diagrams
27
Data Collection
28
Testing
29
Automation: Workflow Models
30
Toolchain Support for Data Provenance
31
System-level Assurance Technology Dynamic assurance
Description:
Explore the challenges and opportunities in developing assurance-based learning-enabled cyber-physical systems for autonomous vehicles in this 41-minute conference talk by Gabor Karsai from Vanderbilt University. Delve into the integration of Learning-Enabled Components (LECs) in Cyber-Physical Systems (CPS) and the complexities of ensuring safety and functionality. Examine formal verification techniques, monitoring technology for assurance, and the formalization of safety case argumentation processes. Discover an engineering process and toolchain for systematic assurance of CPS with LECs, focusing on autonomous vehicles. Learn about reachability analysis of neural networks, closed-loop verification, perception robustness verification, and assurance monitoring techniques such as Inductive Conformal Prediction and anomaly detection. Gain insights into the ALC (Assured Learning-enabled Components) project vision, workflow models, and system-level assurance technology for dynamic assurance in autonomous systems. Read more

Towards Assurance-Based Learning-Enabled Cyber-Physical Systems

Institute for Pure & Applied Mathematics (IPAM)
Add to list