The vast range of challenging real-world scenarios that can be encountered, including rare ‘edge cases’, means that exhaustively testing automated driving systems is an intractable problem.

As a result, advanced sampling techniques are needed to make testing practicable, such as ‘search space optimisation’, where each successive batch of tests is selected based upon a statistical analysis of previous test results such that the test programme iteratively homes in on key areas of the scenario space for maximum efficiency.

• How to use intelligent sampling techniques to iteratively find scenarios that trigger system errors
• How to incorporate different test modalities (e.g. simulation, proving ground) with efficiency and robustness
• How to analyse results to determine when sufficient statistical confidence has been achieved that the system is free of unreasonable risk