Collection and classification of influence parameters for safety effectiveness of ADAS

Virtual scenario-based testing has become an acceptable method for evaluating
safety effectiveness of advanced driver assistance systems (ADAS). Due to the
complexity of the ADAS operating environment, the scenarios that an ADAS could
face are almost infinite. Therefore, it is crucial to find critical scenarios to improve
the efficiency of testing without compromising credibility. One popular method is
to explore the parameterized scenario space using various intelligent search
methods. Selecting parameters to parameterize the scenario space is
particularly important to achieve good coverage and high efficiency. However,
an extensive collection of (relevant) influence parameters is missing, which allows
a thorough consideration when selecting parameters regarding specific scenarios.
In addition, the general importance definition for individual influence parameters
is not provided, regarding the potential influence of their variations on the safety
effectiveness of ADAS, which can also be used as a reference while selecting
parameters. Combining knowledge from different sources (the published
literature, standardized test scenarios, accident analysis, autonomous vehicle
disengagement, accident reports, and specific online surveys), this paper has
summarized, in total, 94 influence parameters, given the general definitions of
importance for 77 influence parameters based on cluster analysis algorithms. The
list of influence parameters provides researchers and system developers a
comprehensive basis for pre-selecting influence parameters for evaluating the
safety effectiveness of ADAS by virtual scenario-based testing and helps check
whether certain influence parameters can be a meaningful extension for the
evaluation.