Short bio: Riccardo Groppo took his MSc degree in Electronic Engineering at the Politecnico of Torino (Torino, Italy). He is the co-founder and CEO of Ideas & Motion, a high-tech company focused on IP development on silicon and design of complex automotive control systems for niche applications. He is the Chairman of Transportation Working Group and Board Vice-Chairman within EPoSS (European Platform on Smart Systems Integration). He is Member of the Technical Committee in some relevant events worldwide (SAE World Congress, AMAA Conference and Smart System Integration Conference). He started his career with Honeywell Bull and then joined Centro Ricerche FIAT (CRF) in 1989, where he was involved in the design of innovative engine/vehicle automotive control systems. He was a member of the CRF team who developed the first automotive Common Rail system for a direct injection Diesel engine. Then he was involved in the design and industrialization of the MultiAir Technology and the Dry Dual Clutch transmission. He has been the Head of the Automotive Electronics Design and Development Dept. at CRF (2002-2013), where he promoted the design of IP building blocks by means of ASIC technology in cooperation with Freescale Semiconductors and Robert BOSCH for the FIAT/Chrysler applications. Those smart drivers are the standard de-facto in automotive powertrain applications, with volumes exceeding 17 Million parts/year. He holds more 31 patents in the field of automotive electronics and embedded systems, most of which are currently in production on passenger cars.

Abstract: AI has seen an explosion in real-world applications in the recent years. For example, it is the backbone of self-driving and connected cars. The design of AI hardware accelerators to support the intensive and memory-hungry AI workloads is an on-going effort aiming at optimizing the energy-area trade-off. This special session will focus on dependability aspects in the design of AI hardware accelerators.  It is often tacitly assumed that neural networks on hardware inherit the remarkable fault tolerance capabilities of the biological brain. This assumption has proven to be false in recent years by a number of fault injection experiments. The three talks will cover reliability assessment and fault tolerance of Artificial Neural Networks and Spiking Neural Networks implemented in hardware, as well as the impact of approximate computing on the fault tolerance capabilities.

Presentations:

• Fault Tolerance of Neural Network Hardware Accelerators for Autonomous Driving
  Adrian Evans (CEA-LETI, Grenoble, France), Lorena Anghel (Grenoble-INP, SPINTEC, Grenoble, France), and Stéphane Burel (CEA-LETI, Grenoble, France)

• Exploiting Approximate Computing for Efficient and Reliable Convolutional Neural Networks
  Alberto Bosio (École Centrale de Lyon, INL, Lyon, France)

• Reliability Assessment and Fault Tolerance of Spiking Neural Network Hardware Accelerators
  Haralampos-G. Stratigopoulos (Sorbonne University, CNRS, LIP6, Paris, France)

Abstract: Automotive Microcontrollers have been becoming very complex System-on-Chips (SoCs). Especially the megatrends Assisted Driving (ADAS) and Automated Driving (AD), but also traditional applications such as power-train steering require ever-increasing functionality. However, these safety-critical environments require zero defect, and the implementation of functional safety measures and the rising complexity poses significant challenges to satisfy these requirements.This special session addresses these challenges and shows potential solutions to overcome them with the help of the EDA industry.

Presentations:

• Automated solutions for safety and security vulnerabilities

Teo Cupaiuolo (Synopsys)

• Functional Safety: an EDA perspective

Alessandra Nardi (Cadence)

• The Zero Defect Goal For Automotive ICs

Lee Harrison (Siemens EDA); Nilanjan Mukherjee (Siemens)

Abstract: High quality demanding products segments like automotive, transportation, and aerospace have been characterized by persistent needs across several years:

• Zero defects, or in general very low defective levels
• Accurate modeling and prediction of product reliability

The sustainability of these objectives is challenged by the relentless demand of higher performances products and the consequent access to higher complexities and advanced technology nodes.
Functional safety standards and requirements aim to grant the usability of products in safety-critical applications and add several requirements whose satisfaction is a key criticality during the development of a new product.

The proposed panel session would like to debate with the experts about how much the available EDA tools are effectively helping to face the described challenges.
As an example, these are suitable questions that anyone in the field may need answering:

• How does EDA help effectively resolve “end-to-end” the traceability of defined requirements? Is this representing a sustainable effort?
• Is DFT effective enough in addressing fault models to reach target quality?
• Is verification/simulation/validation effective respect transient mode?