Navigating the Evolving Landscape of Safety Standards for Machine Learning-based Road Vehicle Functions

Abstract

<div class="section abstract"><div class="htmlview paragraph">ML approaches to solving some of the key perception and decision challenges in automated vehicle functions are maturing at an incredible rate. However, the setbacks experienced during initial attempts at widespread deployment have highlighted the need for a careful consideration of safety during the development and deployment of these functions. To better control the risk associated with this storm of complex functionality, open operating environments, and cutting-edge technology, there is a need for industry consensus on best practices for achieving an acceptable level of safety.</div><div class="htmlview paragraph"><b>Navigating the Evolving Landscape of Safety Standards for Machine Learning-based Road Vehicle Functions</b> provides an overview of standards relevant to the safety of ML-based vehicle functions and serves as guidance for technology providers—including those new to the automotive sector—on how to interpret the evolving standardization landscape. The report also contains practical guidance, along with an example from the perspective of a developer of an ML-based perception function on how to interpret the requirements of these standards.</div><div class="htmlview paragraph"><a href="https://www.sae.org/publications/edge-research-reports" target="_blank">Click here to access the full SAE EDGE</a>^TM<a href="https://www.sae.org/publications/edge-research-reports" target="_blank"> Research Report portfolio.</a></div></div>