Safety of Embedded Artificial Intelligence

Location: Workshop Room II (onsite), Hofburg, Vienna
Date: September 23, 2024
Time: 16:00 – 17:30 CEST

Host

Workshop description 

In 2022, The Autonomous welcomed Infineon Technologies as Lead in Safety of Embedded Artificial Intelligence. As part of the previous Main Event 2023 as well as of the ongoing Expert Circle on Safety of Embedded AI, dedicated solutions of how to use AI safely for trajectory planning and control were created. It was shown that these AI solutions reduce system costs, increase energy efficiency, and improve passenger comfort. 

This workshop invites participants to join the ongoing conversations and discuss potential technical example solutions. The topics to be addressed in this workshop are:    

• We recap results from the last year on how conventional trajectory control methods can be enhanced by AI techniques and increase the tracking accuracy by roughly 50%.  

• We show an AI enhanced MPC for trajectory planning and how reinforcement learning is used to adapt the MPC parameters as well as weights online in order to achieve better energy efficiency. 

• We show a procedural approach based on standards (ISO 8800, ISO 26262 & ISO 21448) including what to consider to enable a safe AI system for trajectory planning and control. 

• Since AI models are data-driven, one first has to ensure the safety of the data based on standards which include: data completeness, integrity and correctness. The latter also tackles the safety of AI models. Additionally, the correct selection of data has to be ensured by the correct definition of the Operational Design Domain (ODD) and instance scenarios. 

• Once in operation, AI systems require continuous oversight to remain within their ODD boundaries. Run time monitors play a critical role in this process, ensuring that the system does not deviate from its defined constraints. 

Ahead of this workshop, you can read more about the  Expert Circle here 

WORKSHOP SPEAKERS 

• Juergen Schaefer, Senior Principal Engineer System Architecture – Infineon 
• Mathias Haberjahn, Senior Manager, AI Software and Applications – Infineon 
• Stephan Baumgaertner, AI expert – Infineon 
• Lukas Marik, AI expert – Infineon 
• Daniel Scholz, AI expert – Infineon 
• Lucas Garcia, PhD, Principal Deep Learning Product Manager – MathWorks