Decision Making of Connected Automated Vehicles at an Unsignalized Roundabout Considering Personalized Driving Behaviours

Chen Lv (吕辰), Peng Hang, Chao Huang, Zhongxu Hu, Yang Xing, Chen Lv

TL;DR

This paper introduces a game-theoretic framework for connected automated vehicles (CAVs) navigating unsignalized roundabouts, specifically accounting for personalized driving styles. By integrating motion prediction with Stackelberg and grand coalition game approaches, the system aims to achieve safe, comfortable, and efficient navigation, demonstrating its effectiveness through simulations.

What Problem Does This Paper Address?

Unsignalized roundabouts present a complex traffic scenario due to the need for multi-vehicle interaction and decision-making without traffic signals. While existing research has explored centralized control or simplified interactions, this paper tackles the challenge of individual CAVs making decisions in a dynamic, multi-lane unsignalized roundabout environment. A key gap identified is the lack of consideration for personalized driving behaviors (e.g., aggressive vs. conservative) which significantly influence human driver interactions and, by extension, should impact CAV decision-making for realistic and accepted autonomous driving.

Methods

The core of the proposed solution is a game-theoretic decision-making framework. This framework incorporates:

  • Motion Prediction: A module using Model Predictive Control (MPC) to predict the future states of CAVs, enhancing the accuracy of the decision-making algorithm.
  • Personalized Driving Styles: Three distinct driving styles are defined: aggressive (prioritizing efficiency), conservative (prioritizing safety), and normal (a balance). These styles are embedded in the payoff functions.
  • Payoff Function Definition: The decision-making payoff function considers safety, comfort, and travel efficiency. Safety is further broken down into longitudinal, lateral, and lane-keeping safety.
  • Game-Theoretic Approaches:
    • Stackelberg Game: This approach models a leader-follower dynamic, where the "leader" (e.g., the host vehicle) can predict the "followers'" (other vehicles) strategies to optimize its own payoff. It aims to guarantee personalized driving.
  • Grand Coalition Game: This is a cooperative approach where all vehicles form a coalition to maximize the collective payoff (system efficiency).
  • MPC Optimization: Both game-theoretic approaches are integrated with MPC to solve the formulated optimization problems and derive optimal decision-making sequences.

Datasets

Not discussed in the paper. The paper focuses on simulation-based testing of the proposed framework.

Key Insights and Ideas

  • Personalized Driving Matters: The study strongly emphasizes that individual driving styles (aggressive, conservative) significantly impact decision-making at roundabouts, influencing merging behavior, speed, and safety margins.
  • Stackelberg vs. Grand Coalition:
    • The Stackelberg game effectively preserves personalized driving, allowing each CAV to pursue its own objectives (safety, efficiency) based on its style.
  • The Grand Coalition game prioritizes overall system efficiency, potentially at the cost of individual personalized driving preferences (e.g., a conservative driver might be pushed to drive faster for system benefit).
  • Trade-offs are Inherent: The research highlights a fundamental trade-off between individual personalized driving and overall system efficiency, which different game-theoretic approaches can balance differently.
  • Roundabout Complexity: The framework is designed for a two-lane roundabout with multi-vehicle interactions, increasing its relevance for complex urban environments beyond simpler intersection scenarios.

Key Results

The paper presents simulation results from three testing cases involving different driving style combinations and scenarios (entering, passing, exiting).

  • Scenario A (Normal, Conservative, Normal, Normal): Conservative driving styles led to safer, more cautious maneuvers, with the host vehicle (HV) merging into the outside lane.
  • Scenario B (Normal, Aggressive, Normal, Normal): An aggressive NV1 forced the HV to merge into the inside lane, demonstrating how aggressive behavior influences others.
  • Scenario C (Aggressive, Aggressive, Normal, Normal): Aggressive driving styles by HV and NV1 led to faster speeds and more assertive maneuvers, but also reduced safety margins (smaller relative distances).
  • General Findings:
    • The Stackelberg game generally resulted in velocities more aligned with individual driving styles compared to the Grand Coalition game.
  • The Grand Coalition game showed improved system velocity (overall traffic efficiency) but potentially sacrificed individual personalization.
  • Aggressive driving styles led to higher accelerations and smaller safety gaps, while conservative styles resulted in greater deceleration and larger safety gaps.
  • The computational time for both SG and GC approaches was found to be efficient enough for real-time applications.

Limitations and Future Work

  • Limitations:
    • The current framework primarily considers CAVs and does not explicitly model interactions with human-driven vehicles, which is a significant aspect of real-world traffic.
  • The study is based on simulations; real-world validation is needed.
  • The computational burden increases with the number of players (vehicles), which could be a challenge for very large scenarios.
  • Future Work:
    • Extend the framework to handle interactions between CAVs and human-driven vehicles.
  • Apply the algorithms to more complex traffic scenarios.
  • Improve the computational efficiency of the solvers and explore deployment on high-performance computing platforms for real-time hardware implementation.
  • Further refine the decision-making cost functions and behavior representation variables for more nuanced scenarios.

Summary updated

Abstract

To improve the safety and efficiency of the intelligent transportation system, particularly in complex urban scenarios, in this paper a game theoretic decision-making framework is designed for connected automated vehicles (CAVs) at unsignalized roundabouts considering their personalized driving behaviours. Within the decision-making framework, a motion prediction module is designed and optimized using model predictive control (MPC) to enhance the effectiveness and accuracy of the decision-making algorithm. Besides, the payoff function of decision making is defined with the consideration of vehicle safety, ride comfort and travel efficiency. Additionally, the constraints of the decision-making problem are constructed. Based on the established decision-making model, Stackelberg game and grand coalition game approaches are adopted to address the decision making of CAVs at an unsignalized roundabout. Three testing cases considering personalized driving behaviours are carried out to verify the performance of the developed decision-making algorithms. The testing results show that the proposed game theoretic decision-making framework is able to make safe and reasonable decisions for CAVs in the complex urban scenarios, validating its feasibility and effectiveness. Stackelberg game approach shows its advantage in guaranteeing personalized driving objectives of individuals, while the grand coalition game approach is advantageous regarding the efficiency improvement of the transportation system.
BibTeX
@article{2021decision,
  author = {Chen Lv (吕辰) and Peng Hang and Chao Huang and Zhongxu Hu and Yang Xing and Chen Lv},
  title = {{Decision Making of Connected Automated Vehicles at an Unsignalized Roundabout Considering Personalized Driving Behaviours}},
  journal = {IEEE Trans. Vehicular Technology},
  year = {2021},
  doi = {10.1109/tvt.2021.3072676},
}
Figure