Two upcoming webinars focus on key aspects of designing the kinds of advanced driver assistance systems (ADAS) that will pave the way for fully autonomous vehicles.
Designing for such safety-critical applications means understanding and learning to work with a new set of constraints that must be applied during the design and verification of an SoC in order for it to achieve the certifications necessary for it to be accepted for use in vehicle systems.
These constraints include meeting the ISO26262 functional safety standard, which defines a set of 'automotive safety integrity levels' (ASILs) that describe the potential risks within an automotive system. These risks are defined by combining measure of the severity of a failure, the probability the failure will result in a safety hazard, and the probability that harm can be avoided when a hazardous condition occurs.
Systems built for use in automotive applications also have to meet stringent reliability and quality management standards, such as AEC Q100 for reliability and TS 16949 for quality management. Meeting the demands of these standards means developing a deep understanding of their philosophy, how that philosophy is reflected in the standards' requirements, and how to meet those requirements in practice.
The first webinar, on 4 April, will consider one key aspect of ADAS - using deep learning and embedded vision to help vehicles understand the dynamic environments in which they are operating.
Early uses of this technology have included object detection, including pedestrian detection and collision avoidance. However, the technology is rapidly developing to the point at which every pixel of a high-resolution video stream can be identified, classified, and used as input to the decision-making algorithms of an ADAS. The webinar will touch on the use of convolutional neural networks and deep-learning accelerator technology to ease the implementation of such sophisticated systems in SoCs.
The second webinar, which will be held 12 April, will focus on the use of pre-certified IP blocks to speed up the development of ADAS SoCs. Using certified IP can help designers reach the point at which they are ready to seek ISO 26262 certification more quickly. If the IP blocks have been developed to meet the demands of AEC Q100 reliability qualification and TS 16949 quality standards, this should also ease the process of meeting stringent automotive industry requirements.
The webinar will discuss best practices for the development and certification of ASIL Ready IP according to ISO 26262, and review the requirements for achieving AEC Q100 reliability and TS 16949 automotive quality.