Taking on the data challenge in ADAS and autonomy

There are many challenges, and a significant proportion of these are related to safety – which is clearly an important matter, as people will be in vehicles over which they have little, if any, direct control.

Testing these
vehicles needs to be adequate to catch any defect and cover all possible driving situations – but how much testing will be required? In 2016, the RAND Corporation estimated that billions of kilometres would have to be driven –
although, more recently, the focus has been on the diversity of driving situations as opposed to billions of highway miles.

Ensuring diversity of miles

Due to regulatory standards, testing is an indispensable facet of all stages of the development
process for autonomous driving (AD) software and advanced driver-assistance systems (ADAS). It is often performed by undertaking many thousands of simulation exercises – often in parallel, to minimise time to market.

To enable this rapid parallel testing, the test
system must be capable of storing petabytes of test data that may reside on servers anywhere around the world. Furthermore, the test results and analysis must be made available to thousands of engineering teams globally...

Read the full article in the August issue of DPA