In recent years, the adoption of Artificial Intelligence (AI) at the edge, particularly in the automotive industry, has grown exponentially. Edge AI, essential for Advanced Driver Assistance Systems (ADAS), enables real-time detection of road hazards, significantly enhancing driver safety. AI accelerators provide the computational power critical to these applications. As these accelerators become integral to automotive systems, ensuring their safety through robust safety mechanisms is crucial. In addition, tailored Electronic Design Automation (EDA) methodologies are essential for designing AI accelerators that meet the stringent safety standards required in the automotive industry.
This Ph.D. research focuses on enhancing the fault tolerance and safety of AI accelerators through innovative EDA methods. The research will involve analysing selected AI accelerators to assess their inherent fault tolerance and identify safety-critical components. The objective is to design hardware accelerators with advanced fault detection and correction mechanisms and to pinpoint areas requiring enhanced fault detection. This project will develop innovative methods to leverage EDA tool flows to efficiently detect and correct faults in AI accelerators. Additionally, existing Cadence tools and flows will be adapted to integrate AI capabilities. The developed concepts and methodologies will be demonstrated within the Cadence Functional Safety tool flow, applying them to real-world case studies.
Specific background:
Requirements:
Note: This PhD project is part of the TIRAMISU European HORIZON MSCA Doctoral Network. The application for DC2.3 must also be submitted via the TIRAMISU website: https://tiramisu-project.eu/vacancies/application-procedure
Benefits we offer you:
And so much more, do not hesitate to contact us.