Game-theoretic Modeling of Traffic in Unsignalized Intersection Network for Autonomous Vehicle Control Verification and Validation

Abstract

For a foreseeable future, autonomous vehicles (AVs) will operate in traffic together with human-driven vehicles. The AV planning and control systems need extensive testing, including early-stage testing in simulations where the interactions among autonomous/human-driven vehicles are represented. Motivated by the need for such simulation tools, we propose a gametheoretic approach to modeling vehicle interactions, in particular, for urban traffic environments with unsignalized intersections. We develop traffic models with heterogeneous (in terms of their driving styles) and interactive vehicles based on our proposed approach, and use them for virtual testing, evaluation, and calibration of AV control systems. For illustration, we consider two AV control approaches, analyze their characteristics and performance based on the simulation results with our developed traffic models, and optimize the parameters of one of them.

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

“…However, only two players, i.e., the ego vehicle and an opponent vehicle, are considered. In [27], the level-k game theory is applied to model multi-vehicle interactions at an unsignalized intersection, and this approach is combined with the receding-horizon optimization and imitation learning to design the decisionmaking framework for CAVs. In general, the game theoretic approaches can be applied not only to handle the decision making of CAVs, but also to simulate the interactive behaviors of intelligent agents [28].…”

Analyzing Multiagent Interactions in Traffic Scenes via Topological Braids

“…Another body of work has focused on tools for testing and validation in realistic settings, leveraging a semantic-level understanding of interactions. Tian et al [42] model traffic at unsignalized intersections using tools from game theory and propose a verification testbed for navigation algorithms. Liebenwein et al [22] propose a framework for safety verification of driving controllers based on compositional and contract-based principles.…”

“…Driver behavior can often be modeled as rational, characterized by risk aversion and efficiency-seeking objectives. Recent work has leveraged these observations in the design of data-driven behavior prediction and planning frameworks [42,7,34,17]. To perform robustly in the real world, such frameworks require large, balanced datasets containing highly diverse behaviors.…”

A Survey on Autonomous Vehicle Control in the Era of Mixed-Autonomy: From Physics-Based to AI-Guided Driving Policy Learning

“…The hierarchical game-theoretic model is tested on two scenarios with merging and overtaking maneuvers: one on a straight empty multi-lane highway with only two-vehicle interaction and one with the presence of a third vehicle (i.e., a truck with a slower moving speed). Li et al (2018a); Tian et al (2018Tian et al ( , 2019 assume two different game structures for HVs and AVs, respectively. Human drivers play a game based on hierarchical reasoning.…”

“…dataset aggregation (DAgger) is one popular technique to mitigate propagation errors of BC by augmenting original training data with expert demonstration for missing states (Ross et al, 2011). Assuming human drivers follow hierarchical reasoning decision-making, Tian et al (2019) employs DAgger to establish a mapping from the ego car's state, all others' state, and the ego car's reasoning level k to the ego car's level k action. To accommodate heterogeneity in human drivers, different HVs are assumed to follow different reasoning levels.…”