Multi-vehicle conflict detection and resolution method based on integrated behavioral decision-making in complex parking scenarios

Abstract

Complex parking environments involve highly coupled paths among multiple intelligent vehicles and correspondingly high conflict densities. In high-conflict-density environments, traditional Conflict-Based Search (CBS) algorithms suffer from redundant node expansion, low computational efficiency, and infeasible path generation. To address these challenges, this paper introduces a multi-vehicle conflict detection and resolution framework that integrates Time Conflict-Based Search (T-CBS) with game-theoretic behavioral decision-making. At the low level, the integration of the directed A* algorithm and the Hybrid A* algorithm enables the generation of paths that closely resemble human driving trajectories while satisfying vehicle kinematic constraints. At the high level, the T-CBS algorithm incorporates a global time variable to track and prioritize the earliest conflict, thereby enabling a chain reaction that can resolve subsequent conflicts. A complete-information static game-theoretic decision mechanism then uses a multi-criteria payoff function to identify which vehicle to constrain, avoiding simultaneous constraints on multiple vehicles. Furthermore, a deterministic equilibrium-selection mechanism is introduced to ensure that a unique executable conflict-resolution decision can still be obtained when multiple pure-strategy equilibria exist or when no pure-strategy equilibrium is available. Simulation results demonstrate that the low-level planner reduces grid usage and computation time compared to Hybrid A* alone. At the high level, the integrated method produces fewer nodes and conflicts while maintaining both driving/parking safety and overall operational efficiency.