Characterizing Lane-Changing Behavior in Mixed Traffic

TL;DR

This study addresses how autonomous and human-driven vehicles interact during lane-changing in mixed traffic. It introduces a data-driven, game-theoretic framework that uses the Waymo Open Motion Dataset to classify cooperative versus defective lane-changing behaviors via clustering, estimate utilities with a quantal response equilibrium, and construct empirical payoff tables. The work then applies evolutionary game theory to simulate how cooperation evolves under varying AV market penetration rates and social contact frequencies, revealing that AVs tend to cooperate more, social dilemmas occur in a minority of states, and cooperation increases over time with repeated interactions. The findings provide a principled approach to quantifying and promoting cooperative lane-changing in mixed-traffic environments, with implications for AV policy, design, and connected-vehicle protocols, while noting limitations such as the lack of AV–AV interaction data and the focus on successful lane changes.

Abstract

Characterizing and understanding lane-changing behavior in the presence of automated vehicles (AVs) is crucial to ensuring safety and efficiency in mixed traffic. Accordingly, this study aims to characterize the interactions between the lane-changing vehicle (active vehicle) and the vehicle directly impacted by the maneuver in the target lane (passive vehicle). Utilizing real-world trajectory data from the Waymo Open Motion Dataset (WOMD), this study explores patterns in lane-changing behavior and provides insight into how these behaviors evolve under different AV market penetration rates (MPRs). In particular, we propose a game-theoretic framework to analyze cooperative and defective behaviors in mixed traffic, applied to the 7,636 observed lane-changing events in the WOMD. First, we utilize k-means clustering to classify vehicles as cooperative or defective, revealing that the proportions of cooperative AVs are higher than those of HDVs in both active and passive roles. Next, we jointly estimate the utilities of active and passive vehicles to model their behaviors using the quantal response equilibrium framework. Empirical payoff tables are then constructed based on these utilities. Using these payoffs, we analyze the presence of social dilemmas and examine the evolution of cooperative behaviors using evolutionary game theory. Our results reveal the presence of social dilemmas in approximately 4% and 11% of lane-changing events for active and passive vehicles, respectively, with most classified as Stag Hunt or Prisoner's Dilemma (Chicken Game rarely observed). Moreover, the Monte Carlo simulation results show that repeated lane-changing interactions consistently lead to increased cooperative behavior over time, regardless of the AV penetration rate.

Figures (6)

• Figure 1: Sample 10-second trajectories of extracted lane-changing (LC) events: (a) HDV only, (b) Waymo as a LC vehicle, (c) Waymo as a lead vehicle, and (d) Waymo as a lag vehicle. The starting points of all vehicles are shown in lighter colors, and the ending points in darker colors. The LC vehicle is shown in red, the lead vehicle in green, and the lag vehicle in blue. The road edge is shown as a black solid line, white solid and dashed lane markings on the road are represented as gray solid and dashed lines in the figure, and the yellow centerline is depicted as a yellow solid line.
• Figure 2: Illustration of defined variables from the Waymo Open Dataset. (a) Definition of lane-changing timing, including the start, lane change, and end points, as well as the lane crossing angle. (b) Definition of speed gain and the time periods before, during, and after the lane change.
• Figure 3: Proportions of cooperative AVs and HDVs across active and passive roles.
• Figure 4: Social dilemmas in lane-changing based on empirical payoffs from the perspective of (a) active and (b) passive vehicles, for each interaction type.
• Figure 5: Results of Monte Carlo simulations after 200 time steps: Sensitivity to (a) interaction neighbor size, (b) noise parameter, (c) market penetration rate, and (d) social contact frequency.