A Cognitive Environment Modeling Approach for Autonomous Vehicles: A Chinese Experience

Abstract

Existing environment modeling approaches and trajectory planning approaches for intelligent vehicles are difficult to adapt to multiple scenarios, as scenarios are diverse and changeable, which may lead to potential risks. This work proposes a cognitive spatial–time environment modeling approach for autonomous vehicles, which models a multi-scenario-adapted spatial–time environment model from a cognitive perspective and transforms the scenario-based trajectory planning problem into a unified spatial–time planning problem. The commonality of multiple typical Chinese road scenarios is analyzed, and a unified spatial–time environment model for multi-scenario adaptation is defined and established. The adaptability and trajectory planning potential of the spatial–time environment model are analyzed, and the planning results are obtained through a hybrid A* algorithm. The simulation results show that the proposal is effective in blurring the boundary between scenarios, allowing a single planning approach to adapt to multiple scenarios and plan optimal trajectories (optimal in both path and speed domains) and introducing more flexibility to the planning.