基于高斯混合-隐马尔可夫模型的驾驶意图识别

打开文本图片集

主题词：自动驾驶 驾驶意图识别高斯混合模型隐马尔可夫模型Frenet坐标系中图分类号：U463.6 文献标志码：A DOI：10.19620/j.cnki.1000-3703.20231201

Driving Intention Recognition Based on Gaussian Mixture-Hidden MarkovModel

Shen Yu1²2，Liu Guanghui²，Ma Xuanpeng1，XuJiawen²，Yan Yuan² （1.School of Information Engineering，Gansu Minzu Normal University，Hezuo 747ooo；2.Schoolof Electronicand Information Engineering，Lanzhou Jiaotong University，Lanzhou ）

【Abstract】To achieve accurate recognition of vehicle driving intentions in highway scenarios，this paper proposes a drivingintentionrecognitionmodelthatcombinesdualreferencelinesintheFrenetcoordinatewith GausianMixture Models （GMMs）andHidden Markov Models （HMMs）.Themodel selectsdrivingdata fromdiferentreference linesintheFrenet coordinatebasedonvehiclepositionasobservedvariables.By integratingtheobservationprobabilitiesoutputbytheGMMat previousand subsequent time steps withthe HMM，the model identifies the vehicles’driving intentionat thecurrentmoment. Theefectivenessof themodelisvalidatedusingtheUS-101datasetfrom NGSIM.Theresultsshow that thedual-referenceline GMM-HMM model achieves recognition accuracies of 93.33% for lane keeping and 92.24% for lane changing，indicating excellent recognition performance.

Key Words：Autonomous driving，Driving intention recognition，Gaussan Mixture Model （GMM），HiddenMarkovModel （HMM），Frenetcoordinate

【引用格式】沈瑜，刘广辉，马翱鹏，等.基于高斯混合-隐马尔可夫模型的驾驶意图识别[J].汽车技术，2025（5）：22-28SHENY，LIUGH，MAXP，etal.Driving IntentionRecognitionBasedonGausian Mixture-HiddenMarkovModel[J].Automobile Technology，2025（5）：22-28.

1前言

道路交通事故原因分析表明，约四分之一的安全事故源于驾驶意图传达不明确[1。（剩余9483字）