Introduction
In modern urban environments, right-turn-on-red (RTOR) intersections represent complex social and physical spaces where multiple road users—motor vehicles, e-bikes, and pedestrians—must negotiate right-of-way simultaneously. While a traditional "social contract" often exists between drivers and pedestrians (where vehicles typically yield to those on foot), the rapid rise of electric bicycles (e-bikes) has disrupted this equilibrium.
This study identifies a critical "dual role" dilemma: e-bikes behave like vulnerable pedestrians when interacting with cars but act as dominant motor vehicles when interacting with pedestrians. This ambiguity creates a "rationality gap," making it difficult for road users to predict behavior and increasing the risk of accidents. This paper quantifies these interactions using a game-theoretic framework to understand the underlying behavioral chaos.
Methodology
The researchers modeled the intersection as a static game where players make simultaneous decisions, such as "yield" versus "not yield" or "cross" versus "stay." To account for the inherent uncertainty and human error in high-stress traffic situations, the study employed the Quantal Response Equilibrium (QRE) model. Unlike the standard Nash Equilibrium, which assumes perfect rationality, QRE introduces a rationality parameter ($\lambda$) to capture the "bounded rationality" of real-world actors.
Data for the model was extracted from 10 hours of surveillance footage across three distinct sites in Ningbo, China, ranging from formal signalized crossroads to informal unsignalized intersections. The team utilized a sophisticated computer vision pipeline to process this data:
- YOLO v8 and ByteTrack algorithms were deployed to extract multi-target trajectories.
- A single-adaptive matrix mapping method converted 2D video coordinates into a projective coordinate system for precise speed and distance calculations.
- Payoff functions were then constructed based on the competing priorities of Efficiency (passing speed) and Safety (collision avoidance).
Key Behavioral Insights
The analysis of 503 valid interactions revealed several critical findings regarding the impact of e-bikes on traffic safety:
- The Rationality Gap: The rationality parameter ($\lambda$) was significantly lower in games involving e-bikes ($\lambda \approx 2.1$ to $4.4$) compared to vehicle-pedestrian games ($\lambda \approx 5.3$). This indicates that the presence of e-bikes leads to more "irrational" or error-prone decision-making by all parties involved.
- E-bikes as "Pseudo-Vehicles": Kinematically, e-bikes behave more like cars than pedestrians, entering the interaction zone from greater distances and maintaining higher average speeds (~15 km/h). However, their legal classification as non-motorized vehicles in China creates a regulatory gray area that complicates right-of-way expectations.
- Broken Social Contracts: In vehicle-pedestrian interactions, the hierarchy is clear. However, e-bikes frequently ignore yielding norms, consistently preferring to "not yield" or "cross" regardless of the other party's actions, which forces other users into defensive or confused states.
- Velocity Pressures: Over 30% of observed e-bikes exceeded 20 km/h, with 13% surpassing the 25 km/h legal limit. These high speeds drastically reduce the decision-making window for other road users, further destabilizing the interaction.
Results and Conflict Analysis
The QRE model successfully converged for all scenarios, validating that efficiency and safety are the primary drivers of traffic behavior. Key statistical results include:
- Conflict Probability ($p_{conflict}$): The likelihood of a "cross/not yield" scenario was highest in vehicle-e-bike interactions (0.418), notably higher than vehicle-pedestrian interactions (0.339).
- Confusion Probability ($p_{confusion}$): The probability of mutual hesitation ("not cross/yield") was significantly higher in e-bike-pedestrian interactions (0.120) than in vehicle-pedestrian interactions (0.099).
Conclusion and Future Work
The study concludes that e-bikes act as "wild cards" that degrade the predictability of urban intersections. While the current model provides a robust framework for understanding these dynamics, future research aims to expand the sample size of e-bike-pedestrian interactions and incorporate demographic factors like age and gender. Ultimately, the researchers suggest that these game-theoretic insights could be integrated into autonomous vehicle systems and smart city infrastructure to better predict and mitigate human-driven traffic conflicts.
Summary updated