Skip to content

Project

HPC telemetry analysis

portrait As compute clusters continue to grow in scale and complexity, the analysis of HPC telemetry becomes increasingly challenging. This project aims to leverage novel learning methods to understand system dynamics and the relationships between telemetry data.

Resilience Evaluation of Autonomous Vehicle Models

portrait The technology behind self-driving vehicles has greatly improved in the past decade, mainly due to rapid advances in deep neural networks (DNNs), making it necessary to detect faults in these safety-critical systems that could cause safety hazards or accidents.

  • L2 ADAS System Openpilot and L4 system LBC are both vulnerable to faults.
  • Existing protection mechanisms are helpful but insufficient for preventing all hazards.
  • TGFI efficiently finds fault sites that lead to hazards.