Responsibility Sensitive Safety Dangerous Situation (RSS-DS)#
Description#
The Responsibility Sensitive Safety (RSS) framework is designed to formally guarantee safety during an automated vehicle’s drive. It was developed to reflect a sound interpretation of law that leads to a efficient and verifiable AV behavior [Shalev-Shwartz2017]. To approach this goal, RSS states a set of mathematical rules.
For this, the safe lateral and longitudinal distances \(d_\mathit{min}^\mathit{lat}\) and \(d_\mathit{min}^\mathit{long}\) are formalized, depending on the current road geometry. The metric RSS-DS for the identification of a dangerous situation \(S\) with a set of actors \(\mathcal{A}\) is defined as
Note that to determine \(d_\mathit{min}^\mathit{lat}\) and \(d_\mathit{min}^\mathit{long}\), different prediction models are utilized to estimate which distances are classified as safe, e.g. for intersections, highways, and unstructured roads.
Note that RSS has been shown to not consider certain edge cases, e.g. during braking maneuvers and on varying road surfaces and slopes, as well as the issue of perception uncertainty [Koopman2019].
An extension of RSS-DS measures the temporal extent to which the ego was not able to mitigate the dangerous situation [Jesenski2020]. In accident research, a similar concept of classifying situations as safe and unsafe depending on longitudinal stopping distances was introduced as the Stopping Distance Index (SDI) [Oh2006]. In turn, the SDI is partially based on the idea of the Potential Index for Collision with Urgen Deceleration (PICUD) [Uno2002], both comparing the stopping distances of the lead and following vehicle under emergency braking.
Properties#
Run-time capability#
Yes
Target values#
Not necessary
Subject type#
Road vehicles (esp. suitable for automated vehicles, but also possible to evaluate human drivers)
Scenario type#
RSS ODD (also suited for urban, unstructured scenarios)
Inputs#
\(d^\mathit{lat}(A_1, A_i)\) and \(d^\mathit{long}(A_1, A_i)\) for all \(A_1 \neq A_i\), response time \(\rho\) and other inputs required to predict \(d^\mathit{lat}_\mathit{min}\) and \(d^\mathit{long}_\mathit{min}\)
Output scale#
\(\{0,1\}\), number, nominal scale
Reliability#
Medium, the nominal nature of the metric’s scale can lead to fluctuations if vehicles exist close the boundaries of the safe distancesMedium, the nominal nature of the metric’s scale can lead to fluctuations if vehicles exist close the boundaries of the safe distances
Validity#
High, depending mainly on the validity of the safe distance definition of the scenario (e.g. highways or unstructured roads) [Chai2019]
Sensitivity#
High, due to over-approximation of safe space [Chai2019], although reduced for edge cases [Koopman2019]
Specificity#
Medium, as not every violation of a safe space directly implies high criticality [Chai2019], but depends on the definition of the safe distances
Prediction model#
Time window#
Depends on model for \(d_\mathit{min}\) prediction, for single lane roads \(\rho + \mathit{TTB}\)
Time mode#
Linear time