基于神经网络的自动驾驶模型及其虚拟实现

文章提出了一种基于人工神经网络的自动驾驶控制模型,并利用计算机虚拟技术模拟实现车辆的运行环境及其运行行为,并对自动驾驶控制模型进行了测试。试验表明这种自动驾驶模型能有效地指挥车辆的驾驶。

基于元胞自动机的自动驾驶交通流仿真研究

将Arnab Bose自动驾驶模型与经典的Na Sch模型相结合,提出一个适用于自动驾驶的单车道元胞自动机交通流模型。利用MATLAB语言对自动驾驶车辆运行情况进行仿真分析,发现模拟结果较好地呈现了自动驾驶环境下的交通流特征。另外,还发现在自动驾驶中设置不同的车头时距对道路通行能力以及拥堵会产生很大的影响,车头时距为0.5 s时的通行能力约是车头时距为3 s的通行能力的4倍,当车头时距从4 s减小为1 s时交通拥堵可降低约95%,可以作为一种缓解拥堵的策略。

针对多种交通标识分类模型的攻击及鲁棒性分析

为了保证自动驾驶系统的安全性符合预期功能安全的要求,有必要对交通标识牌分类模型实施对抗攻击。基于此,采用两种基于图形学噪声的攻击方法对交通标识牌分类模型实施对抗攻击。同时,在4种交通标识牌分类模型上开展对抗攻击实验,通过分析普遍逃逸率(Universal Evasion Rate,UER)的值来分析此方法的可行性和神经网络的鲁棒性。

基于MM-STConv的端到端自动驾驶行为决策模型

针对现有端到端自动驾驶模型输入数据类型单一导致预测精确度低的问题,选取RGB图像、深度图像和车辆历史连续运动状态序列作为多模态输入,并利用语义信息构建一种基于时空卷积的多模态多任务(Multimodal Multitask of Spatial-temporal Convolution,MM-STConv)端到端自动驾驶行为决策模型,得到速度和转向多任务预测参量。首先,通过不同复杂度的卷积神经网络提取场景空间位置特征,构建空间特征提取子网络,准确解析场景目标空间特征及语义信息;其次,通过长短期记忆网络(LSTM)编码-解码结构捕捉场景时间上、下文特征,构建时间特征提取子网络,理解并记忆场景时间序列信息;最后,采用硬参数共享方式构建多任务预测子网络,输出速度和转向角的预测值,实现对车辆的行为预测。基于AirSim自动驾驶仿真平台采集虚拟场景数据,以98200帧虚拟图像及对应的车辆速度和转向角标签作为训练集,历经10000次训练周期、6h训练时长后,利用真实驾驶场景数据集BDD100K进行模型的测试与验证工作。研究结果表明:MMSTConv模型的训练误差为0.1305,预测精确度达到83.6%,在多种真实驾驶场景中预测效果较好;与现有其他主流模型相比,该模型综合场景空间信息与时间序列信息,在预测车辆速度和转向角方面具有明显的优势,可提升模型的预测精度、稳定性和泛化能力。

基于注意力机制的多模态自动驾驶行为决策模型

针对现有端到端自动驾驶模型未考虑驾驶场景中不同区域的重要性和不同语义类别之间的关系而导致预测准确率低的问题,受驾驶人注意力机制和现有端到端自动驾驶模型的启发,充分考虑驾驶场景的动态变化、驾驶场景的语义信息和深度信息对驾驶行为决策的影响,以连续多帧驾驶场景的RGB图像为输入,构建一种基于注意力机制的多模态自动驾驶行为预测模型,实现对方向盘转角和车速的准确预测。首先,通过语义分割模型和单目深度估计模型分别获取RGB图像的语义图像和深度图像;其次,为剔除与驾驶行为决策无关信息,以神经科学和空间抑制理论为基础,设计一种拟人化注意力机制作为能量函数来计算驾驶场景中不同区域的重要度;为学习语义图像中与驾驶行为决策最为相关类别之间的关系,采用图注意力网络(Graph Attention Network, GAT)对驾驶场景的语义图像进行特征提取;然后,以保留RGB特征为原则对提取的驾驶场景的图像特征、语义特征和深度特征进行融合,采用卷积长短期记忆网络(Convolutional Long Short Term Memory, ConvLSTM)实现融合特征在连续多帧之间的传递,进而实现下一帧驾驶场景对应驾驶行为的预测;最后,与其他模型的对比试验、消融试验、泛化试验和特征可视化试验来充分验证所提出自动驾驶行为预测模型的性能。试验结果表明:与其他驾驶行为预测模型相比,所提出模型的训练误差为0.021 2,预测准确率为86.97%,均方误差为0.031 5,其驾驶行为的预测性能优于其他模型;连续多帧的语义图像和深度图像、拟人化注意力机制和面向语义特征提取的GAT有助于提升驾驶行为预测的性能;该模型具有较好的泛化能力,其做出驾驶行为预测所依赖的特征与经验丰富的驾驶人所关注的特征基本一致。

Intelligent back-looking distance driver model and stability analysis for connected and automated vehicles

The connected and automated vehicles(CAVs)technologies provide more information to drivers in the car-following(CF)process.Unlike the human-driven vehicles(HVs),which only considers information in front,the CAVs circumstance allows them to obtain information in front and behind,enhancing vehicles perception ability.This paper proposes an intelligent back-looking distance driver model(IBDM)considering the desired distance of the following vehicle in homogeneous CAVs environment.Based on intelligent driver model(IDM),the IBDM integrates behind information of vehicles as a control term.The stability condition against a small perturbation is analyzed using linear stability theory in the homogeneous traffic flow.To validate the theoretical analysis,simulations are carried out on a single lane under the open boundary condition,and compared with the IDM not considering the following vehicle and the extended IDM considering the information of vehicle preceding and next preceding.Six scenarios are designed to evaluate the results under different disturbance strength,disturbance location,and initial platoon space distance.The results reveal that the IBDM has an advantage over IDM and the extended IDM in control of CAVs car-following process in maintaining string stability,and the stability improves by increasing the proportion of the new item.

Anomaly detection method based on kinematics model and nonholonomic constraint of vehicle

A method used to detect anomaly and estimate the state of vehicle in driving was proposed.The kinematics model of the vehicle was constructed and nonholonomic constraint conditions were added,which refer to that once the vehicle encounters the faults that could not be controlled,the constraint conditions are violated.Estimation equations of the velocity errors of the vehicle were given out to estimate the velocity errors of side and forward.So the stability of the whole vehicle could be judged by the velocity errors of the vehicle.Conclusions were validated through the vehicle experiment.This method is based on GPS/INS integrated navigation system,and can provide foundation for fault detections in unmanned autonomous vehicles.

Modeling and gender difference analysis of acceptance of autonomous driving technology

In order to deeply analyze the differences in the acceptance of autonomous driving technology among different gender groups,a multiple indicators and multiple causes model was constructed by integrating a technology acceptance model and theory of planned behavior to comprehensively reveal the gender differences in the influence mechanisms of subjective and objective factors.The analysis is based on data collected from Chinese urban residents.Among objective factors,age has a significant negative impact on women's perceived behavior control and a significant positive impact on perceived ease of use.Education has a significant positive impact on men's perceived behavior control,and has a strong positive impact on women's perceived usefulness(PU).For men,income and education are found to have strong positive impacts on perceived behavior control.Among subjective factors,perceived ease of use(PEU)has the greatest influence on women's behavior intention,and it is the only influential factor for women's intention to use autonomous driving technology,with an influence coefficient of 0.72.The influencing path of men's intention to use autonomous driving technology is more complex.It is not only directly affected by the significant and positive joint effects of attitude and PU,but also indirectly affected by perceived behavior controls,subjective norms,and PEU.

A Cellular Automaton Model for Heterogeneous and Incosistent Driver Behavior in Urban Traffic

In this paper a cellular automaton model is proposed to describe driver behavior at a single-lane urban roundabout. Driver behavior has been considered as heterogeneous and inconsistent. Most traffic papers in the literature just discussed heterogeneous driver behavior, to our best knowledge. Two truncated Caussian distributions are used to model heterogeneous and inconsistent driver behavior, respectively. The physical meanings of two truncated distributions are indicated. This method may help enhance a better understanding of driver behavior at roundabout traffic, and even possibly provide references for roundabout design and management.