Research On Path Planning Method Of Urban Autonomous Driving Based On High-Definition Map

Autonomous vehicles are an important component of intelligent transportation and one of the strategic development directions for the future.As a product of the integration and development of cutting-edge technologies such as intelligent transportation,automobile manufacturing,high-definition maps,and perception,autonomous driving has been elevated to the level of a national strategy and has become a hot topic in both academic and industrial circles around the world.Autonomous driving technology is developing rapidly in the automotive industry and has made significant breakthroughs in known areas such as highways and closed roads,as well as in good weather and signal conditions.However,urban traffic presents various road scenes with different features,such as structured roads(overpasses,loops),semi-structured roads(internal roads),and unstructured roads(parking lots),as well as phenomena such as urban canyons,vegetation obstruction,construction barriers,and illegal parking.When facing complex urban traffic environments,stable,safe,and efficient autonomous driving remains a challenge.In order to operate smoothly in urban environments,autonomous vehicles require real-time collaboration among multiple modules such as perception,localization,mapping,decision-making,path planning,and control,with the coordination between mapping and path planning being particularly critical.The accuracy and richness of information provided by traditional electronic navigation maps are insufficient to meet the needs of autonomous driving vehicles.Path planning requires data support from high-definition maps to achieve urban autonomous driving.However,there are currently various high-definition map models,and a lack of uniform standards leads to inconsistent road representation for different road scenes in urban areas.Real-time processing of complex road scenes in cities and rapid seamless switching between different scenes are also research challenges for path planning methods.To address these issues,this article proposes a high-definition mapbased autonomous driving path planning method,providing technical support for path planning in urban autonomous driving.The specific research content of this article includes the following aspects:1)Introduction of research background and significance,main problems,and research approach.Firstly,the research background and significance of path planning methods for autonomous driving vehicles are introduced.Secondly,the current domestic and international development status of autonomous driving vehicles is shared.Then,the main research approach is explained in response to the problems of inconsistent high-definition map representation,poor adaptability of path planning,and low passage rate.Finally,the main research content and structural arrangement of this paper are summarized.2)Summarize the problem of urban autonomous driving path planning.Firstly,the method of dividing urban road scenes for autonomous driving was introduced,including structured roads,semi-structured roads and unstructured roads.Secondly,the composition framework of autonomous driving path planning system was analyzed,including global path planning and local path planning.Finally,the definition of the problem of autonomous driving path planning was elaborated,including vehicle kinematic model,definition and terminology of path planning problem,and extension strategy of path planning.3)In response to the issues of insufficient prior information provided by traditional electronic navigation maps and inconsistent representation of different road scenes in high-definition maps,a high-definition map-based safe driving area generation method is proposed.Firstly,the composition framework of the safe driving area is analyzed,which includes two parts: collision risk map and reference trajectory map.Secondly,the generation method and expression form of the collision risk map are introduced.Finally,the generation method of the reference trajectory map in different road scenes is explained.This method unifies the representation method of complex and changeable road scenes in cities,provides consistent guidance information for autonomous driving path planning,and effectively solves the problem of inconsistent representation in high-definition maps.4)In response to the issue of a single path planning method being unable to handle complex and changeable urban road scenes,an adaptive path planning model based on the safe driving area is proposed.Firstly,the composition structure of path planning is introduced.Secondly,urban road scenes are divided into structured,semistructured,and unstructured roads.Then,an adaptive path planning model based on the safe driving area is proposed for these three different urban road scenes.Furthermore,an improved SST-based global path planning algorithm is proposed for these three different road scenes,providing directly executable reference trajectories and planning boundaries for local path planning.Finally,three global path planning experiments are designed for the three road scenes,and the advantages and disadvantages of the traditional electronic navigation map-based and the safe driving area-based global path planning methods are compared.5)In response to the issues of slow convergence speed and low computational efficiency of local path planning algorithms in complex urban road scenes,an HDMRRT based on the safe driving area is proposed.Firstly,the motion model of autonomous driving vehicles is introduced,and the definition,expansion strategy,and constraint conditions of the path planning problem are defined.Secondly,the algorithm framework and advantages and disadvantages of the RRT and RRT* algorithms are analyzed.Then,an RRT algorithm based on high-definition maps(HDM-RRT)is proposed,which includes the algorithm framework of HDM-RRT,a Gaussian sampling method based on collision risk maps,a cost equation based on collision risk coefficients,a two-layer node optimization method based on Clothoid distance,and a trajectory optimization method based on multiple constraint conditions.The optimization effect of the HDM-RRT method is summarized.Finally,a comparison experiment is designed between the Gaussian sampling method based on collision risk maps and the traditional random distribution-based sampling method and Gaussian distribution-based sampling method,analyzing the sampling efficiency of the three methods.6)The proposed high-definition map-based autonomous driving path planning method was experimentally validated using the Wuhan University "Tu Zhong Hao" autonomous driving car.The experiment was conducted on the internal roads of Wuhan University campus.A route containing various types of roads,including structured,semi-structured,and unstructured roads,was designed in the experimental area.The route included scenarios such as straight roads,irregular bends,intersections,Tjunctions,U-turns,and parking lots.Through the analysis,comparison,and accuracy evaluation of the experimental results,it was verified that the proposed high-definition map-based path planning method can adaptively handle autonomous driving problems in different road scenes in cities.It can provide efficient,real-time,and safe drivable trajectories for autonomous driving cars within limited computing resources and time.