自动驾驶的安全性是国内外智能汽车领域广泛专注的焦点，预期功能安全是自动驾驶安全

的重要组成部分，与雨雪雾恶劣天气、传感器遮挡等高风险长尾场景强相关，长尾难题是目前

制约自动驾驶产业化的关键问题。项目针对“如何实现长尾场景下的自动驾驶安全”这一问

题，对鲁棒感知模型、不确定性度量、长尾测试数据生成三方面关键技术进行闭环研究：一.

针对长尾场景下的传感器数据特征，采用全局信息增强策略，提出视觉时序融合自监督、点云

网格形状增强方法及对比预训练多模态鲁棒融合模型；二.基于特征共形预测思想，提出AI模

型的不确定性量化方法，准确且实时实现面向长尾场景的模型不确定性度量；三.研究适用于

自动驾驶的生成式人工智能方法，结合风险触发机制合成长尾场景，生成真实度高、覆盖率高

的仿真测试数据，丰富危险场景测试用例库。项目拟解决长尾难题的部分关键技术问题，提

高自动驾驶的安全性，促进自动驾驶的社会认可度及产业化进程。

主持/参与科研项目: 

国家自然科学基金项目:  面向自动驾驶环境认知的态势评估功能模型及实现方法

国家自然科学基金重点项目：工业生产系统全要素融合的交互学习和协同决策

国家自然科学基金重点项目：基于信息融合的异构智能网联车协同决策与调度

科技部国家重点研发计划新能源汽车专项：自动驾驶电动汽车评价理论研究- AEV预期功能安全风险评估理论研究

科技部国家重点研发计划新能源汽车专项：智能电动汽车全状态参数估计、复杂环境感知与多源信息融合

中央高校科研专项资金重大国际合作预研项目：智能卡系统的攻击及防御关键技术研究

近几年论文： 

[1] LooselyCoupled Stereo VINS Based on Point-Line Features Tracking With Feedback Loops.L Zhang, W Ye, J Yan, H Zhang, J Betz, H Yin. IEEE Transactions on VehicularTechnology, 2024. 

[2] EnhanceAdversarial Robustness via Geodesic Distance. J Yan, H Yin, Z Zhao, W Ge, JZhang. IEEE Transactions on Artificial Intelligence, 2024. 

[3] Robustobject detection for autonomous driving based on semi-supervised learning. WChen, J Yan, W Huang, W Ge, H Liu, H Yin. Security and Safety, 2024. 

[4] RandomNetwork Distillation Based Deep Reinforcement Learning for AGV Path Planning.Huilin Yin, Shengkai Su, Yinjia Lin, Pengju Zhen,Karin Festl, Daniel Watzenig.35th IEEE Intelligent Vehicles Symposium (IEEE IV 2024). 2024. 

[5]SA-Attack:Speed-adaptive stealthy adversarial attack on trajectory prediction. Huilin Yin，Jiaxiang Li，PengjuZhen，Jun Yan.35th IEEE Intelligent Vehicles Symposium (IEEE IV 2024). 2024. 

[6] CSANet:Cuboid-Wise Shape Augmentation 3D Object Detector for Occluded Targets. J. Lin,W. Ge, G. Rigoll, and H. Yin. IEEE Signal Processing Letters, 2024. 

[7] DP-VINS:Dynamics Adaptive Plane-Based Visual-Inertial SLAM for Autonomous Vehicles.Linchuan Zhang, Wei Ye, Johannes Betz, and Huilin Yin. IEEE Transactions onInstrumentation and Measurement. 2024. 

[8] Exploring aesthetic procedural noisefor crafting model-agnostic universal adversarial perturbations. J Yan, H Yin, W Ge, L Liu, Displays 79, 2023. 

[9] Attack Detection for IntelligentVehicles via CAN-Bus: A Lightweight Image Network Approach. S Gao, L Zhang, L He, X Deng, H Yin, H Zhang, IEEE Transactions onVehicular Technology, 2023. 

[10] FSFNet: Foreground score-aware fusionfor 3D object detector under unfavorable conditions. J Lin, H Yin, J Yan, K Jian, Y Lu, W Ge, H Zhang, G Rigoll, IEEE SensorsJournal 23 (14), 15988-16001, 2023. 

[11] DSENet: a deep sub-ensemblesconvolutional neural network for robust semantic segmentation. H Yin, X Xu, Q Meng, International Conference on Cloud Computing,Performance Computing, and Deep Learning (CCPCDL), 2023. 

[12] An Adversarial Attack on SalientRegions of Traffic Sign. J Yan, H Yin, BYe, W Ge, H Zhang, G Rigoll, Automotive Innovation, 1-14, 2023. 

[13] A Survey of Vehicle TrajectoryPrediction Based on Deep-Learning.H Yin, Y Wen, J Li, International Conference on NeuralNetworks, Information and Communication Engineering (NNICE), 2023. 

[14] Ground-optimizedSLAM with Hierarchical Loop Closure Detection in Large-scale Environment. H Yin, MSun, L Zhang, J Yan, J Betz, IEEE International Conference on Intelligent Transportation Systems(ITSC), 2023. 

[15] Multi-ObjectTracking with Object Candidate Fusion for Camera and LiDAR data. H Yin, Y Lu, JLin, M Schratter, D Watzenig, IEEE InternationalConference on Intelligent Transportation Systems (ITSC), 2023. 

[16] AGVPath Planning Using Curiosity-driven Deep Reinforcement Learning. H Yin, Y Lin,J Yan, Q Meng, K Festl, L Schichler, D Watzenig, IEEE CASE, 2023. 

[17]Waveletregularization benefits adversarial training. JYan, H Yin, Z Zhao, W Ge, HZhang, G Rigoll, Information Sciences, 1-20, 2023. 

[18] Multi-agent reinforcement learningfor cooperative lane changing of connected and autonomous vehicles in mixedtraffic. W Zhou, D Chen, J Yan, Z Li, H Yin,W Ge, Autonomous Intelligent Systems 2 (1), 2022. 

[19] Improved 3d object detector undersnowfall weather condition based on lidar point cloud. J Lin, H Yin, J Yan, W Ge, H Zhang, G Rigoll, IEEE Sensors Journal 22(16), 16276-16292, 2022. 

[20] On adversarial robustness of semanticsegmentation models for automated driving.H Yin, R Wang, B Liu, J Yan, IEEE Intelligent Vehicles Symposium (IV), 867-873,2022. 

[21] Curriculum Defense: An EffectiveAdversarial Training Method. H Yin, X Deng, JYan, Chinese Control Conference (CCC), 7399-7406, 2022. 

[22] Review on Uncertainty Estimation inDeep-Learning-Based Environment Perception of Intelligent Vehicles. H Yin, Z Chen, J Yan, G Rigoll, SAE Technical Paper, 2022.