Quantum Convolutional Neural Networks

Tyre Noise Recognition
Safety challenges

  • Usage of ML
  • Open context environment
  • Compliance with laws and standards

Domain Analysis

  • False Positives (FP) as critical case
  • Domain analysis identified triggering events potentially leading to FPs
  • Known triggering events: Decomposition of known physical properties from tyre/road interaction
  • Identified triggering events lead to safety requirements for the system

Verfication & Validation

Objectives

  • Confirmation of assumptions made during system design and safety assurance
  • Evaluation of the function with regard to known triggering events
  • Evaluation of the potential for unknown triggering events
  • Evaluation of the resilience of the function with regard to residual unknown triggering events

Via

  • Analysis (strength & weaknesses)
  • Simulation (noise generation)
  • Structured testing (specific corner-cases)
  • Field tests (public road)

Summary

  • First use of the IKS Assurance Case approach for Safe AI in an industry project
  • Complexity of TNR should allow for building convincing Assurance Case
  • Increased visibility through publication at SAFECOMP 2021

Get in touch

Secure your AI Models with us
We offer deep insights into Artificial Intelligence

QR code with link to s.fhg.de/iks-qc
publica.fraunhofer.de