Research On Behavior And Motion Planning Methods Of Scenario Vehicle In Automated Driving Virtual Testing

Virtual simulation is a key technology for the innovative development of automated driving vehicles.In automated driving simulation test environments,the representation of the dynamic driving environment around the ego vehicle is of particular importance.The scenario vehicles in this area should possess diverse characteristics and a high degree of credible bidirectional interaction with the ego vehicle and others.In current mainstream automated driving virtual simulation tools,scenario vehicles often decelerate immediately when encountering other traffic participants to prevent interaction,resulting in a driving environment that is significantly different from the real traffic environment.Establishing scenario vehicles with human-like and differentiated characteristics,and thereby constructing an effective dynamic driving environment for the ego vehicle’s near-range area,has become a major challenge faced by researchers worldwide.The current simulation lacks a realistic dynamic driving environment around the ego vehicle.To a large extent,this limitation stems from the lack of human-like behavior in scenario vehicles,particularly their decision-making and motion planning.The existing rule-based methods treat different factors on the same level,neglecting considerations of driver psychological needs and hierarchical cognitive processes.They can only respond to traffic events but cannot actively interact with other traffic participants.The pre-defined trajectory methods predetermine the driving routes before simulation,overlooking the constantly changing boundary conditions during driving and the differences in vehicle performance and driver preferences,making it challenging to represent a diverse dynamic driving environment.Furthermore,scenario vehicles cannot intuitively present the decision-making process,lacking consideration of interpretability in behavior selection,making it difficult to analyze the causes of accidents.In addressing the aforementioned issues,this paper explores and proposes a behavioral planning model for scenario vehicles driven by driver psychological demands and with high computational efficiency.The method generates diverse driving behaviors by simulating the variation of driver psychological demands and the hierarchical decisionmaking process,which intuitively presents a more human-like decision-making process and differentiated driving styles.By theoretically deriving optimal control solutions for behaviors,it achieves instant and rapid motion trajectory planning considering vehicle maneuverability and comfort preferences.In addition,this paper studies a real-time simulator for the dynamic driving environment in the near-range of the ego vehicle,aiming to provide an effective dynamic driving environment for automated driving virtual testing.The research in this paper primarily covers the following aspects:In the aspect of scenario vehicle behavioral decision-making,addressing the issues related to insufficient consideration of driver psychological needs and cognitive processes,this paper explores a hierarchical decision-making approach that takes into account the psychological demands of drivers.Initially,based on Psi cognitive psychology,the paper establishes an urge-motive-behavior hierarchical decision-making model,breaking down the decision-making process into urge activation,motivation generation,and behavior decision three levels,providing a human-like description of the decision-making mechanism of scenario vehicles.In the urge activation section,considering the interpretability issues of scenario vehicle behavior,the paper analyzes and proposes five demands influencing driver operations(namely safety,dominance,achievement,order,and relatedness)based on psychology.Corresponding urge assessment methods are provided,offering a basis for the interpretability of scenario vehicle behavior.In the motivation generation section,the paper establishes that the activated urge with higher priority serves as the dominant motivation,assigning higher decision priority.In the behavior decision section,addressing the problem of a limited range of behaviors for scenario vehicles,three categories of behavior sets are determined: changing speed,changing lanes,and maintaining the current state.Each category includes specific behaviors,which ensures the diversity of behavior selection.Corresponding behavior tree models are established for each motivation,enabling scenario vehicles to determine actions that align with the motivation from alternative behaviors based on the surrounding environment,thereby enhancing the interactivity of scenario vehicles.In the aspect of scenario vehicle motion planning,addressing the issues of neglecting the performance of different vehicles and driver preferences and overlooking the continuously changing boundary conditions during travel,this paper explores a motion planning analytical method that considers maneuverability and comfort.By avoiding numerical solutions involving nonlinear constraints such as vehicle dynamics,this method achieves feasible and efficient trajectory planning,providing assurance for the large-scale deployment of scenario vehicles.The executable behaviors proposed in the previous chapter are merged based on the motion process,specifying seven distinct basic behaviors that scenario vehicles need to execute,which effectively reduces the number of motion planning problems.Each behavior is decomposed into main motion that reflect the differences in vehicle maneuverability and driver comfort,along with auxiliary motion that ensure the feasibility of the motion and complement the main motion to accomplish the entire behavior.By constraining the auxiliary motion based on the maximum speed of the dominant motion,the feasibility of the trajectory is ensured.For the dominant motion of each basic behavior,multiple unconstrained optimal control problems are established and solved using the Pontryagin’s minimum principle.The auxiliary motion for each basic behavior is solved using the closed-form solution method of Minimum Jerk.By combining the dominant and auxiliary motions,and considering different vehicle maneuverability and driver comfort preferences,the rapid solution of feasible trajectories is achieved.The third aspect of the research focuses on a real-time simulator for dynamic driving environment in the near-range vicinity of the ego vehicle.Firstly,addressing the inadequacy of existing automated driving virtual simulation tools in simulating the cognitive psychology of scenario vehicles and facilitating real-time trajectory planning based on the surrounding environment,a lightweight scenario vehicle model framework is designed for batch simulation,integrating perceptual,planning,and execution functionalities.This framework incorporates psychologically-informed behavioral decision-making methods and motion planning techniques considering maneuverability and comfort proposed in this paper.Within the perceptual module of this framework,a responsibility-sensitive safety boundary computation method is proposed for dynamic entities in the driving environment,enabling scenario vehicles to perceive dynamic entities within the surrounding safety zone.Additionally,a vehicle detection method with variable visual range based on speed is introduced to perceive vehicles on the same and adjacent lanes.Secondly,addressing the issue of redundant retrieval and processing of static environmental information due to the large number of scenario vehicles in simulation,a structured model of road and intersection entities containing attributes such as road structure and logical connectivity is established,facilitating efficient representation of complex traffic scenarios.To address the difficulty in road information retrieval,a duallayered lane-segment road entity modeling method is proposed for rapid retrieval of road entity information.Furthermore,to streamline the computation of lateral positions in multi-lane road segments,a lane width representation based on the road centerline is suggested for swift lateral position determination.Finally,a graphical real-time simulator for the ego vehicle’s near-range dynamic driving environment is developed,offering an effective environment for automated driving systems.Moreover,it serves as a simulation environment for validating the methods proposed in this paper.In the end,the established scenario vehicle model and dynamic simulation driving environment were validated.Firstly,the proposed hierarchical decision-making method for driving behavior was validated using selected simple typical scenarios.Secondly,the motion planning and computational efficiency of various actions were validated,and the ability of the proposed motion planning method to generate diverse trajectories was demonstrated by testing with speed-changing actions.Furthermore,the key technologies for dynamic driving environment studied in this paper,specifically the modeling of static entities in a dynamic driving environment,were validated.Finally,three typical complex traffic scenarios were designed using the simulation platform built in this paper,validating the effectiveness of the established scenario vehicle model and assessing CPU and memory usage during simulation.The simulation results demonstrate that the proposed model can handle various traffic scenarios.It intuitively presents the motivation generation process and behavior decision-making process,exhibiting certain interactivity and diverse driving styles.The model also has high computational efficiency and low memory usage.