基于避险临界点的行车危险识别与制动策略研究
2025-08-11朱彤张卓张贙兴赵云飞李晓虎
主题词:车辆安全实时风险高危场景识别制动策略避险临界模型中图分类号:U461.91 文献标志码:A DOI:10.19620/j.cnki.1000-3703.20240896
Research on Driving Hazard Identification and Braking Strategy Based on Collision Evasion Point
Zhu Tong',Zhang Zhuo',ZhangYunxing',Zhao Yunfei¹,Li Xiaohu² (1.School of Transportation Engineering,Chang'an University,Xi'an71Oo64;2.ChinaAutomotive Technologyand Research Center, Tianjin 300300)
【Abstract]To quantifydriving risk anddevelopasafe braking strategy,this paper introduces theconcept of Collision Evasion Point (CEP)andbuilds mathematical models for straight-driving and turning scenarios.Using theCEP,ariskrepresentationindexisdefinedtoquantifydrivingrisk.Moreover,116accidentcasesfromChinaIn-DepthAcidentStudy (CIDAS)databaseareclasiiedandtheriskrepresentationindexisapliedtoidentifyhigh-iskcases.Finalldyamic brakingstrategy basedonthebraking-timeindicatorisproposed.Testresultsshowthat,acrossvarious high-riskscenarios, theproposedriskrepresentationindexoutperformsTime-to-Colision(TTC)basedstrategyinidentifyingscene-levelrisk, while thebraking strategyachievesmorereasonable braking timesandsmoother speed profiles,thereby beteravoiding collisions.
Key words: Vehicle safety,Real-time risk,High-risk scenario identification,Braking strategy, Critical risk-avoidancemodel
【引用格式】朱彤,张卓,张兴,等.基于避险临界点的行车危险识别与制动策略研究[J].汽车技术,2025(7):23-30. ZHU T,ZHANG Z,ZHANG YX,et al.Research on Driving Hazard Identification and Braking Strategy Based on Colision EvasionPoint[J].Automobile Technology,2025(7): 23-30.
1前言
近年来,我国道路交通安全形势不容乐观,及时识别行车风险并触发制动对于降低事故率、减轻事故损伤至关重要[1-2]。虽然现阶段车辆已配备车道偏离预警系统、高级驾驶辅助系统等主动安全技术,但风险评估与制动决策模块的可靠性仍有待提升[3-4]。
准确地表征风险是自动驾驶汽车安全性能提升的基础与关键5。现有研究多从不同角度考虑行车风险,通过安全距离指标、距碰撞发生时间指标及驾驶员主观认知判断指标等评价指标表征风险并制定了一系列碰撞预测策略。张宇丰等基于碰撞时间(TimeToCollision,TTC)与安全距离模型协同,对车辆进行控制以避免碰撞;杨贺博等考虑自车车速,优化了TTC值及制动控制策略;郭祥靖等[基于反向传播神经网络对TTC进行预测,并以此制定控制策略;赵林峰等[12]提出了考虑车辆运动状态和路面附着系数等因素的改进安全距离模型。
然而,现有方法通过当前时刻车辆的时空风险程度表征运行风险,包括车辆行驶的空间、时间相对位置以及驾驶员主观风险认知能力等,未考虑当前环境中车辆的风险规避能力,故需综合考虑车辆时空接近程度及避险能力、环境要素等方面,设计表征车辆当前的行车风险的指标。同时,自动驾驶测试场景从单一的纵向安全控制测评,如前车静止[13]、前车慢行[4等,逐步发展到汽车转向场景。……
