Modeling the Dynamics of Driver’s Dilemma Zone Perception Using Machine Learning Methods for Safer Intersection Control

Abbas, Montasir; Machiani, Sahar Ghanipoor; Garvey, Philip M.; Farkas, Andrew; Lord-Attivor, Rene · 2014 · ROSA P / Mid-Atlantic Universities Transportation Center

archive: archived pipeline: cataloged verified

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

Summary

This study addresses the safety challenges associated with the "dilemma zone" (DZ) at signalized intersections, defined as the area where drivers face uncertainty regarding whether to stop or proceed upon the onset of a yellow light. While existing literature extensively covers static factors influencing DZ decisions, a critical gap remains in understanding the dynamic nature of driver perception and how experience alters decision-making. The research aims to identify significant factors affecting driver behavior, investigate whether drivers learn from their experiences, and develop more accurate behavioral models using machine learning techniques to improve intersection control and crash prevention. The methodology employed a multi-phase approach involving a driver survey, a driving simulator experiment, and agent-based modeling. First, a survey administered across Virginia, Pennsylvania, and Maryland collected data from 1,213 participants to identify significant factors influencing DZ decisions. Second, a driving simulator study utilized an Adaptive Randomized Incomplete Block Split-plot (ARIBS) design to examine the learning aspect of driver behavior. This experiment involved 34 participants and tested specific learning hypotheses by adapting scenarios based on individual driver reactions. Third, the researchers applied an actor-critic reinforcement learning algorithm combined with fuzzy logic to model dynamic driver behavior using the simulator data. This approach allowed for the calibration and updating of fuzzy rule policies to capture the effects of driver learning. The findings revealed nine significant factors influencing driver decisions: speed, distance to the intersection, presence of red-light cameras or police, pavement conditions, presence of leading or following vehicles, familiarity with the intersection, and traffic density. Notably, the proportion of drivers citing these factors varied significantly across the three states. The simulator results demonstrated that drivers do not maintain static decision-making patterns; for two out of three learning hypotheses, driver behavior changed after exposure to specific treatments, confirming that drivers learn from their experiences. Furthermore, the agent-based reinforcement learning model showed a close match between simulated driver actions and actual simulator outputs, indicating its effectiveness in capturing dynamic perception. The significance of this research lies in its validation of the dynamic nature of driver decision-making in dilemma zones. The study concludes that traditional static models are insufficient for accurately representing driver behavior. Instead, agent-based modeling and simulation techniques, particularly those incorporating reinforcement learning, provide a more accurate representation of how drivers adapt their behavior based on experience. These findings support the recommendation that future assessments of dilemma zone mitigation strategies and intersection safety designs should utilize agent-based models to account for the learning capabilities of drivers, thereby enhancing the effectiveness of traffic control systems and crash prevention efforts.

Key finding

Drivers' dilemma-zone stop/go decisions depend on nine significant factors identified in a 1,213-respondent survey; simulator evidence shows behavior changes with experience (2/3 learning hypotheses confirmed), and an actor-critic fuzzy RL model matched simulator actions with R²=0.72.

Methodology

mixed_methods

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. Discovered via bulk_ingest_rosap on 2026-05-23 (5 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

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

• gap acceptance
• rail grade crossings
• mental model of traffic
• driver vru interaction
• situational awareness
• traffic density

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).

• Empirical Findings: behavioral performance data
• Theoretical Contribution: computational model, theory or model