A strategy for updating high definition maps in autonomous driving
Compared to traditional electronic navigation maps,high definition maps encompass not only highly accurate geometric information of roadways but also rich attribute data,such as the number of lanes,driving directions,and lane widths.This information aids autonomous vehicles in making informed navigation decisions and ensuring driving safety in complex environments.However,the predominant reliance on specialized collection methods for updating high definition maps is hampered by high costs and low update efficiency,which limits their widespread application.This paper proposes an update strategy for high definition maps,informed by their unique informational characteristics and grounded in a four-layer intelligent high definition map model.This strategy is developed by exploring the differences in update information types,the frequency of detection updates,and the existence of update-driven information across various map information types.It integrates both specialized and crowdsourced updating methods to enhance efficiency.Building on this foundation,the feasibility of the proposed update strategy for different specific application scenarios is validated.Taking Dongcheng District of Beijing as the experimental area,the study designs various experimental scenarios that simulate both extensive changes across the entire road network and localized changes within specific sections of the network.Additionally,the study considers situations with varying frequencies of information changes.In this experiment,information change points randomly distributed across the road network are classified as low-frequency changes(e.g.,changes in roadway information),while points that experience persistent changes over a short time frame are considered high-frequency changes(e.g.,changes in traffic incident information).Subsequently,the study employs both periodic and real-time detection methods to monitor and update changing information across different scenarios,calculating the time,costs,and accuracy required for updating all detected information.Finally,an entropy weight method is utilized to comprehensively evaluate three update indicators.The results indicate that periodic detection,conducted through high-precision equipment on specialized surveying vehicles,is more effective for extensive low-frequency changes in terrestrial information.Conversely,real-time detection effectively captures and updates the latest roadway or traffic information through crowdsourced vehicles in scenarios involving widespread high-frequency changes.Additionally,real-time detection demonstrates flexibility in responding to localized changes in map information,while periodic detection proves effective for monitoring information changes in specific areas,such as minor incident data in remote regions(e.g.,traffic volume information).This experiment illustrates that different updating methods possess distinct advantages and disadvantages across various scenarios,underscoring the necessity to select appropriate update modes based on the type of map information and the frequency of change detection.Furthermore,the proposed map update strategy provides optimal decision-making for updating terrestrial information within high definition maps,contributing to improved update efficiency and cost-effectiveness,thus facilitating the rational allocation and effective utilization of resources.
NETL
NSTL
万方数据
