Managing Driving Modes in Automated Driving Systems

Abstract

Current technology is unable to produce massively deployable, fully automated vehicles that do not require human intervention. Given that such limitations are projected to persist for decades, scenarios requiring a driver to assume control of a semiautomated vehicle, and vice versa, will remain a feature of modern roadways for the foreseeable future. Herein, we adopt a comprehensive perspective of this problem by simultaneously considering operational design domain supervision, driver and environment monitoring, trajectory planning, and driver-intervention performance assessment. More specifically, we develop a modeling framework for each of the aforementioned functions by leveraging decision analysis and Bayesian forecasting. Utilizing this framework, a suite of algorithms is subsequently proposed for driving-mode management and early warning emission, according to a management by exception principle. The efficacy of the developed methods is illustrated and examined via a simulated case study.

Some statistical challenges in automated driving systems

“…Several performance‐related measures (e.g., the total attained utility or the number of crashes) were calculated to assess the strategy's effectiveness. Detailed results of these experiments can be found in Ríos Insua et al, 16 which demonstrate that the proposed strategy consistently performs in a reasonable manner.…”

“…Finally, $$$ p\left({\boldsymbol{\theta}}_1|{\boldsymbol{Y}}_1\right) $$$, $$$ p\left({\boldsymbol{X}}_1|{\boldsymbol{\theta}}_1\right) $$$, and $$$ p\left({\boldsymbol{Y}}_1\right) $$$ respectively represent the distribution over driver states given the initial environmental variables, the distribution over the sensor measurements given the initial driver state, and the distribution over the environmental variables. Based on these, $$$ p\left({\boldsymbol{Y}}_{t+1}|{\boldsymbol{Y}}_t\right) $$$ can be derived via the recursive computations in Ríos Insua et al; 16 the resultant probabilities may then be used to inform driving‐mode and warning‐issuance decisions.…”

“…Finally, p(𝜽 1 |Y 1 ), p(X 1 |𝜽 1 ), and p(Y 1 ) respectively represent the distribution over driver states given the initial environmental variables, the distribution over the sensor measurements given the initial driver state, and the distribution over the environmental variables. Based on these, p(Y t+1 |Y t ) can be derived via the recursive computations in Ríos Insua et al; 16 the resultant probabilities may then be used to inform driving-mode and warning-issuance decisions.…”

“…This ADS management framework is qualitatively and holistically summarized in Algorithm 1. Detailed subroutines for the processes described previously are provided in Ríos Insua et al 16 …”

“…The remainder of this section sketches a statistics‐based approach for managing RtI decisions, presents initial empirical results, and discusses open problems. For a more detailed examination of this problem and the methods discussed herein, we refer an interested reader to Ríos Insua et al 16 …”