Development and Application of a Public Service Platform for Autonomous Driving Testing
Autonomous driving has garnered considerable attention across various sectors,including government,industry,and academia.Scientific testing and evaluation play pivotal roles in advancing autonomous driving technologies.However,in the R&D realm,several common challenges have emerged,including disagreements among development entities regarding algorithm effectiveness,low coverage of testing scenarios,high entry barriers to testing tools,slow testing iteration speeds,and a lack of standardized evaluation criteria.These challenges have led to a growing demand for the development of public service platforms for online testing in the field of autonomous driving.This study introduced a technical framework for the construction of an autonomous driving testing public service platform known as OnSite.OnSite is characterized by its high scenario coverage,lightweight testing deployment,accelerated testing processes,and standardized evaluation criteria.To further meet the requirements of scenario-based testing in autonomous driving,innovative features such as automatic scenario generation,accelerated testing of critical scenarios,and bidirectional interaction testing between the vehicle under test and the background traffic flow were proposed as unique functionalities of the platform.By analyzing the results of competitions and evaluations hosted on the OnSite platform,this study assessed the weaknesses in autonomous driving planning and decision-making algorithm development and offered insights into fundamental research.Finally,a five-stage development plan was presented,featuring characteristics such as motion-planning-oriented testing,full-stack algorithm-oriented testing,virtual-real fusion testing,collaborative driving testing,and development-testing integration.The OnSite platform provides a"ubiquitous"service for autonomous driving testing,facilitating the transition of theoretical advancements from"shelves"to practical application while overcoming the problem of translating research achievements into real-world implementation.
traffic engineeringautonomous drivingpublic service platformtesting and evalua-tionscenario generationaccelerated testing
NETL
NSTL
万方数据
