A Bayesian framework for preventive assistance at road intersections

Armand, Alexandre; Filliat, David; Ibanez-Guzman, Javier · 2016 · Crossref

DOI: 10.1109/ivs.2016.7535531

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the limitation of current Advanced Driving Assistance Systems (ADAS), which typically provide "curative" assistance (e.g., automatic braking or warnings) only when a collision is imminent. The authors propose a Bayesian framework designed to detect risk situations early enough to offer "preventive" assistance, such as advice, allowing drivers sufficient time to react comfortably. The research is motivated by the high frequency of accidents at road intersections, where human error is a primary factor, and the need for systems that can distinguish between appropriate levels of intervention based on the timing and severity of the risk. The study employs a Dynamic Bayesian Network to model the interaction between vehicle state, driver actuation, and context. The framework defines variables for the true and observed physical states of the vehicle (speed, pose, pedal states), as well as the driver’s intended and expected maneuvers. A key component is the modeling of driver individuality; the authors use Gaussian Processes to learn personalized velocity profiles that represent how specific drivers typically decelerate when approaching a stop intersection. The network calculates the probability of the driver’s intention to stop or react by comparing their current behavior against these learned patterns and expected maneuvers. The framework then evaluates the pertinence of three assistance types—automatic actuation, warning, and advice—based on time and physical constraints, such as whether the required deceleration is too harsh for advice or too mild for automatic braking. The framework was applied to a case study involving a vehicle approaching a stop intersection, using data recorded under controlled conditions. The results demonstrate that the Bayesian model can coherently detect risk situations and identify the most appropriate type of assistance for the specific context. By leveraging learned driver patterns and real-time state estimation, the system successfully distinguishes between scenarios requiring preventive advice versus those necessitating curative warnings or automatic braking. The experiments validate the model's ability to infer risk earlier than conventional collision prediction methods, which often rely on short-horizon trajectory predictions. The significance of this work lies in its shift from reactive to proactive safety assistance. By integrating driver-specific behavioral models and contextual awareness, the proposed framework enables ADAS to provide assistance that aligns with the driver’s reaction time and comfort levels. This approach aims to reduce driver discomfort associated with late-stage interventions and potentially prevent accidents by addressing risks before they become imminent. The study highlights the potential of Bayesian methods in creating more nuanced and personalized driving assistance systems.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• telematics crash prediction
• anticipation

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Theoretical Contribution: computational model