Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Human-Based Risk Model for Improved Driver Support in Interactive Driving Scenarios

Tim Puphal, Benedict Flade, Matti Krüger, Ryohei Hirano, Akihito Kimata

TL;DR

This work tackles the problem of human-based driver support by incorporating sensed driver information into the risk model. The authors combine driver perception—captured as notice, forecast, and inference errors—and driver personalization—via driver types and a risk factor—into a unified framework that augments traditional vehicle-state risk models. The method integrates Perceived Risk Maps with a Gaussian-plus-survival risk formulation and a personalization-weighted warning signal, yielding earlier warnings (e.g., up to $1.55$ s for defensive drivers) and reduced warning errors (up to $28 ext{%}$ for confident drivers) across six interactive driving scenarios in simulation. While promising, the study is limited to simulator results with assumed perfect driver-state estimation; future work includes real-vehicle testing and user studies to assess robustness to sensing noise and personalization mis-specification, aiming to close the simulation-to-reality gap.

Abstract

This paper addresses the problem of human-based driver support. Nowadays, driver support systems help users to operate safely in many driving situations. Nevertheless, these systems do not fully use the rich information that is available from sensing the human driver. In this paper, we therefore present a human-based risk model that uses driver information for improved driver support. In contrast to state of the art, our proposed risk model combines a) the current driver perception based on driver errors, such as the driver overlooking another vehicle (i.e., notice error), and b) driver personalization, such as the driver being defensive or confident. In extensive simulations of multiple interactive driving scenarios, we show that our novel human-based risk model achieves earlier warning times and reduced warning errors compared to a baseline risk model not using human driver information.

Human-Based Risk Model for Improved Driver Support in Interactive Driving Scenarios

TL;DR

This work tackles the problem of human-based driver support by incorporating sensed driver information into the risk model. The authors combine driver perception—captured as notice, forecast, and inference errors—and driver personalization—via driver types and a risk factor—into a unified framework that augments traditional vehicle-state risk models. The method integrates Perceived Risk Maps with a Gaussian-plus-survival risk formulation and a personalization-weighted warning signal, yielding earlier warnings (e.g., up to s for defensive drivers) and reduced warning errors (up to for confident drivers) across six interactive driving scenarios in simulation. While promising, the study is limited to simulator results with assumed perfect driver-state estimation; future work includes real-vehicle testing and user studies to assess robustness to sensing noise and personalization mis-specification, aiming to close the simulation-to-reality gap.

Abstract

This paper addresses the problem of human-based driver support. Nowadays, driver support systems help users to operate safely in many driving situations. Nevertheless, these systems do not fully use the rich information that is available from sensing the human driver. In this paper, we therefore present a human-based risk model that uses driver information for improved driver support. In contrast to state of the art, our proposed risk model combines a) the current driver perception based on driver errors, such as the driver overlooking another vehicle (i.e., notice error), and b) driver personalization, such as the driver being defensive or confident. In extensive simulations of multiple interactive driving scenarios, we show that our novel human-based risk model achieves earlier warning times and reduced warning errors compared to a baseline risk model not using human driver information.
Paper Structure (14 sections, 9 equations, 11 figures, 2 tables)

This paper contains 14 sections, 9 equations, 11 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Contribution
  4. Risk Model
  5. Risk Maps
  6. Human-Based Risk Model
  7. Perceived Risk Maps
  8. Risk Factor
  9. Parametrization
  10. Experiments
  11. Scenario Variation
  12. Simulation Results
  13. Discussion
  14. Conclusion and Outlook

Figures (11)

  • Figure 1: The image shows a driving situation example of a dynamic lane change in which human factors can change the preferred diver support. In this paper, we present a human-based risk model that combines a) the driver perception based on driver errors (i.e., notice error, forecast error and inference error) and b) driver personalization based on driver types (e.g., defensive and confident).
  • Figure 2: Common risk models do not consider the human driver for driver support. The risk model uses only the vehicle states in the driving situation and generates a warning signal.
  • Figure 3: Risk Maps for behavior planning. The risk model can be used in a cost function to find an optimal behavior.
  • Figure 4: The proposed human-based risk model. Perceived Risk Maps includes the driver perception using driver errors and a risk factor is used for driver personalization. This allows the risk model to include the human driver for driver support.
  • Figure 5: Interactive driving scenarios which are analyzed in the experiments of this paper. The driver of the vehicle that is supported in the scenarios is colored green. We show that the proposed human-based risk model allows to improve driver support.
  • ...and 6 more figures