WO2021103625A1 - 一种基于前车与自车互动的短期车速工况实时预测方法 - Google Patents

一种基于前车与自车互动的短期车速工况实时预测方法 Download PDF

Info

Publication number
WO2021103625A1
WO2021103625A1 PCT/CN2020/105463 CN2020105463W WO2021103625A1 WO 2021103625 A1 WO2021103625 A1 WO 2021103625A1 CN 2020105463 W CN2020105463 W CN 2020105463W WO 2021103625 A1 WO2021103625 A1 WO 2021103625A1
Authority
WO
WIPO (PCT)
Prior art keywords
vehicle
vehicle speed
future
speed
historical
Prior art date
Application number
PCT/CN2020/105463
Other languages
English (en)
French (fr)
Inventor
孙超
李军求
孙逢春
郭婷婷
励夏
孙海迪
Original Assignee
北京理工大学
北京首科能源技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京理工大学, 北京首科能源技术有限公司 filed Critical 北京理工大学
Publication of WO2021103625A1 publication Critical patent/WO2021103625A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G08SIGNALLING
    • G08GTRAFFIC CONTROL SYSTEMS
    • G08G1/00Traffic control systems for road vehicles
    • G08G1/01Detecting movement of traffic to be counted or controlled
    • G08G1/052Detecting movement of traffic to be counted or controlled with provision for determining speed or overspeed
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/23Clustering techniques
    • G06F18/232Non-hierarchical techniques
    • G06F18/2321Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • GPHYSICS
    • G08SIGNALLING
    • G08GTRAFFIC CONTROL SYSTEMS
    • G08G1/00Traffic control systems for road vehicles
    • G08G1/01Detecting movement of traffic to be counted or controlled
    • G08G1/0104Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125Traffic data processing
    • G08G1/0129Traffic data processing for creating historical data or processing based on historical data

Definitions

  • the invention relates to the field of vehicle speed condition prediction, in particular to a short-term vehicle speed condition real-time prediction method based on the interaction between the preceding vehicle and the self-vehicle.
  • the purpose of the present invention is to provide a method and system for real-time prediction of short-term vehicle speed conditions based on the interaction between the preceding vehicle and its own vehicle, which can greatly improve the accuracy of future short-term vehicle speed prediction and apply it to energy management strategies to improve Fuel economy.
  • a real-time prediction method for short-term vehicle speed conditions based on the interaction between the preceding vehicle and the self-vehicle including:
  • the valid data includes the effective measured data during the operation of the own vehicle; the instantaneous speed of the vehicle and the operation of the preceding vehicle Valid measured data at the time of the instantaneous speed of the vehicle and the instantaneous distance between the self-vehicle and the preceding vehicle during operation;
  • the future speed prediction model includes a historical basis The prediction model of vehicle speed and the prediction model based on historical distance between vehicles;
  • the future vehicle speed is the vehicle speed at the next moment;
  • the short-term vehicle speed is the vehicle speed in the next time period;
  • the optimal torque distribution strategy is calculated according to the torque demand.
  • the acquiring the historical speed information of the own vehicle, the historical speed information of the preceding vehicle, and the distance information between the own vehicle and the preceding vehicle, and extracting valid data specifically includes:
  • the historical speed information of the own vehicle when different drivers are driving on different road conditions is obtained, stored in the vehicle speed database to form the first sample operating condition, and the first sample operating condition is obtained from the first sample Extract the effective actual measured data of the vehicle running at the time t of each sample operating condition in the operating conditions, the instantaneous speed v 0 (t) of the vehicle;
  • the millimeter-wave radar is used to continuously obtain the vehicle distance information between the vehicle and the preceding vehicle, store it in the historical vehicle distance database to form a third sample operating condition, and extract the vehicle from the third sample operating condition.
  • the constructing a future vehicle speed prediction model according to the historical speed information of the own vehicle, the historical speed information of the preceding vehicle, the distance information between the own vehicle and the preceding vehicle, and the effective data specifically includes:
  • the radial basis function artificial neural network is selected as the nonlinear prediction function to predict the future vehicle speed of the self-vehicle, and the future vehicle speed prediction model based on the radial basis function artificial neural network is constructed; the future vehicle speed prediction model of the radial basis function artificial neural network Including input layer, hidden layer and output layer;
  • the input layer is the input layer
  • the driver behavior Q When constructing a model based on historical vehicle speed information, the driver behavior Q, the historical vehicle speed of the vehicle in the previous time period H h seconds and the historical vehicle speed of the preceding vehicle in the previous time period H s seconds are selected as
  • the input of the future vehicle speed prediction model is defined as:
  • N in is the prediction model input
  • V k and V n are the current vehicle speeds of the vehicle and the preceding vehicle, respectively
  • V k-1 and V n-1 are the vehicle speeds of the vehicle and the preceding vehicle at the previous time, respectively
  • S n is The distance between the preceding vehicle and the vehicle at the current moment
  • S n-1 is the distance between the preceding vehicle and the vehicle at the previous moment
  • H h , H s , H r and H f are all positive integers
  • the hidden layer is the hidden layer
  • the number of neurons O is selected according to the training accuracy requirements of the neural network, and the activation function of the hidden layer is:
  • a 1 exp(-
  • 2 /2b 2 ), n Wa 0 +b;
  • O is a positive integer
  • a 1 and a 0 are the neuron output of the current layer and the previous layer respectively
  • n is the cumulative output
  • c is the center of the neuron node
  • b is the spread width of the neuron radial basis function
  • W is Weights
  • the output layer is the output layer
  • H p is the predicted future vehicle speed vector length
  • the offline training of the constructed future vehicle speed prediction model to determine the trained future vehicle speed prediction model specifically includes:
  • the radial basis function artificial neural network is a g-h-m connection mode, that is, there are g inputs, h hidden layers and m outputs;
  • the gradient descent method is used to adjust the weight ⁇ ij from the hidden layer to the output layer of the network until the network error E ⁇ ends; the network error is expressed by the mean square error, and the expression is as follows:
  • E represents the network error
  • the y(x p ) table is the expected output corresponding to the input
  • N is the number of the input x p.
  • the online prediction of the future vehicle speed of the self-vehicle according to the trained future vehicle speed prediction model specifically includes:
  • the future vehicle speed prediction model after training is a model based on historical vehicle speed
  • the historical vehicle speed in the previous time period H h seconds and the historical vehicle speed data of the previous vehicle in the previous time period H s seconds are used to compare the vehicle Predict the future vehicle speed in H p seconds in the next time period;
  • the future vehicle speed prediction model after training is a model based on historical distance between vehicles , the historical vehicle speed in the previous time period H r seconds and the previous vehicle and the historical vehicle in the previous time period H f seconds of the vehicle are used.
  • the distance data predicts the future vehicle speed of the own vehicle in the next time period H p seconds;
  • the ways to predict the future speed of the vehicle include:
  • the real-time vehicle speed data in the previous time period H r seconds is continuously obtained, and the millimeter wave radar is used to continuously obtain the real-time vehicle speed data of the preceding vehicle and the own vehicle in the previous time period H f seconds.
  • the distance data is combined with the driver’s style to form a neural network input vector to realize the prediction of the future speed of the self-vehicle.
  • the adaptive learning of the trained future vehicle speed prediction model according to the future vehicle speed to determine the adaptive learning future vehicle speed prediction model specifically includes:
  • each cycle is one week, one month or one year ;
  • the future vehicle speed prediction model after training is used to predict the vehicle speed, and an adaptive learning future vehicle speed prediction model is determined.
  • the calculation method of the torque demand is as follows:
  • m is the mass of the car
  • g is the acceleration due to gravity
  • T out is the output torque
  • T break braking torque R wheel is the wheel radius
  • is the road gradient
  • C r is the road resistance coefficient
  • C d is the wind resistance coefficient
  • A is The windward area of the vehicle.
  • the calculation of an optimal torque distribution strategy based on the torque demand based on the dynamic programming algorithm specifically includes:
  • ⁇ k+1 T( ⁇ k ,v k );
  • ⁇ k+1 is the state transition equation
  • T( ⁇ k ,v k ) is the torque demand when the state variable is ⁇ k and the vehicle speed is v k;
  • the objective function is separable and satisfies the recurrence relationship
  • r( ⁇ k ,v k ) represents the cost function corresponding to the state variable ⁇ k and the vehicle speed v k ;
  • r * ( ⁇ k ) represents the optimal cost function when the state variable is ⁇ k ;
  • r * ( ⁇ k+1 ) represents the optimal cost function when the state variable is ⁇ k+1;
  • the terminal conditions are:
  • the optimal solution is to calculate the torque demand T( ⁇ k , v k ) corresponding to the optimal value function after the k value is obtained;
  • the optimal torque distribution strategy under a given initial state is sequentially calculated; the optimal torque distribution strategy includes optimal engine torque and optimal motor torque control variable.
  • the present invention Compared with the prior art, the present invention has the advantages that: the present invention provides a real-time prediction of short-term vehicle speed based on the interactive mode of the preceding vehicle and self-vehicle, using the Internet of Vehicles technology, from the perspective of the human-vehicle-environment system, Through neural network online learning, a method for short-term vehicle speed prediction is proposed, which comprehensively considers vehicle state parameters, driver's driving style, and vehicle state parameters ahead, which improves the accuracy of vehicle speed prediction. And apply the proposed short-term vehicle speed prediction method to the fuel consumption control strategy to obtain better fuel economy.
  • Figure 1 is a block diagram of the system principle of the present invention
  • Figure 2 is a schematic diagram of a hardware device in an embodiment.
  • a method for real-time prediction of short-term vehicle speed conditions based on the interaction between the preceding vehicle and its own vehicle is characterized by the following steps:
  • the equipment that needs to be used includes, but is not limited to, an on-board automatic diagnosis system (OBD), an on-board unit (OBU), a drive test unit (RSU) in a dedicated short-range communication (DSRC) system, and a millimeter wave radar.
  • OBD on-board automatic diagnosis system
  • OBU on-board unit
  • RSU drive test unit
  • DSRC dedicated short-range communication
  • millimeter wave radar can obtain the distance information of the front and rear vehicles.
  • the hardware equipment is shown in Figure 2; step S1 specifically includes:
  • the effective measured data of the running of front and rear vehicles and the time t of each sample working condition are extracted from the third sample working condition of the historical vehicle distance database.
  • the artificial neural network can be selected as various artificial neural network algorithms such as back propagation neural network, layer recurrent neural network, radial basis function neural network, etc.
  • the radial basis function artificial neural network is selected as the model Construct.
  • This step constructs an artificial neural network's future vehicle speed prediction model based on historical vehicle speed or historical vehicle distance information, and specifically includes the following steps:
  • the radial basis function artificial neural network is composed of input layer, hidden layer, and output layer;
  • the input of the network speed prediction model is defined as:
  • N in is the prediction model input
  • V k and V n are the current vehicle speeds of the vehicle and the preceding vehicle, respectively
  • V k-1 and V n-1 are the vehicle speeds of the vehicle and the preceding vehicle at the previous time, respectively
  • S n is The distance between the preceding vehicle and the vehicle at the current moment
  • S n-1 is the distance between the preceding vehicle and the vehicle at the previous moment
  • H h , H s , H r and H f are all positive integers
  • the number of neurons O is selected according to the training accuracy requirements of the neural network.
  • the activation function of the hidden layer is:
  • a 1 exp(-
  • 2 /2b 2 ), n Wa 0 +b;
  • O is a positive integer
  • a 1 and a 0 are the neuron output of the current layer and the previous layer respectively
  • n is the cumulative output
  • c is the center of the neuron node
  • b is the spread width of the neuron radial basis function
  • W is Weights
  • H p is the length of the predicted future vehicle speed vector, that is, the length of the neural network model output vector; suppose f m and f n are nonlinear functions predicted by the neural network ;
  • step S301 Construct input vector parameters and output vector parameters according to step S2, and then input the input parameter vector and output parameter vector into the radial basis function artificial neural network model to form training samples for offline training, and establish a stable radial basis function artificial nerve Network structure
  • the input vector parameters are:
  • the output vector parameters are:
  • the input vector parameters are:
  • the output vector parameters are:
  • the core is to solve the hidden layer basis function center, the variance of the basis function, and the weight of the hidden layer unit to the output unit, so that the jth in the RBF neural network is obtained.
  • the output is expressed as:
  • P is the total number of samples
  • ci is the center of the hidden layer node of the network
  • 2 is the Euclidean norm
  • ⁇ i is the width of the basis function
  • ⁇ ij is the connection weight from the hidden layer to the output layer
  • y j is the actual output of the j-th output node of the neural network corresponding to the input sample;
  • the gradient descent method is used to adjust the weight ⁇ ij from the hidden layer to the output layer of the network until the network error E ⁇ , end; the network error is expressed by the mean square error, and the expression is as follows:
  • E represents the network error
  • the y(x p ) table is the expected output corresponding to the input
  • P is the total number of samples
  • N is the number of the input x p.
  • step S3 If the model obtained in step S3 is a model based on historical vehicle speed, the historical vehicle speed in the previous time period H h seconds and the historical vehicle speed data in the previous vehicle previous time period H s seconds are used to calculate the next time of the vehicle Predict the future vehicle speed within H p seconds;
  • step S3 If the model obtained in step S3 is a model based on historical vehicle distance, use the historical vehicle speed in the previous time period H r seconds and the historical vehicle distance data of the preceding vehicle and the previous vehicle in the previous time period H f seconds. Predict the future speed of the vehicle within H p seconds in the next time period;
  • the ways to predict the future speed of the vehicle include:
  • the vehicle speed acquisition system continuously obtains real-time vehicle speed data in the previous time period H h seconds, and continuously obtains the previous vehicle’s previous time period H s based on the Internet of Vehicles technology (optional "V2V" special short-range communication equipment) Real-time vehicle speed data within seconds, combined with the driver’s style to form a neural network input vector to realize the prediction of the future vehicle speed of the vehicle;
  • V2V Internet of Vehicles technology
  • the real-time vehicle speed data of the vehicle in the previous time period H r seconds is continuously obtained, and the millimeter wave radar is used to continuously obtain the real-time vehicle distance data of the preceding vehicle and the vehicle in the previous time period H f seconds.
  • the neural network input vector is formed by fusing the driver's style to realize the prediction of the future speed of the self-vehicle.
  • S501 Collect the speed data of the vehicle in front and the vehicle ahead or the distance data of the vehicle in front and the vehicle in the previous week, and update the sample vehicle speed or vehicle distance database; each cycle is one week, one month or one year;
  • step S503 Use the neural network model obtained in step S502 to predict vehicle speed;
  • step S504. Return to 501 after collecting new sample data, and repeat the complete process of step S5.
  • the torque demand is calculated as follows:
  • m is the vehicle mass
  • T out is the output torque
  • T break is the braking torque
  • R wheel is the wheel radius
  • is the road gradient
  • C r is the road resistance coefficient
  • C d is the wind resistance coefficient
  • A is the windward area of the vehicle.
  • ⁇ k+1 T( ⁇ k ,v k )
  • the present invention is based on the interactive mode of the preceding vehicle and the self-vehicle, and uses the Internet of Vehicles technology to study the realization of short-term vehicle speed real-time prediction from the perspective of the human-vehicle-environment system.
  • the vehicle's future short-term vehicle speed prediction method based on vehicle state parameters, driver's driving style, and front vehicle state parameters improves the accuracy of vehicle speed prediction. And apply the proposed short-term vehicle speed prediction method to the fuel consumption control strategy to obtain better fuel economy.

Abstract

一种基于前车与自车互动的短期车速工况实时预测方法,包括以下步骤:获取自车和前车历史车速、车距信息,并提取有效数据;构建基于人工神经网络的未来车速预测模型;对构建的未来车速预测模型进行离线训练;在线预测自车的未来车速;实现车速预测神经网络的自适应学习;根据预测的短期车速计算扭矩需求;根据转矩需求和动态规划算法计算最优转矩分配。该方法运用人工神经网络方法对自车短期车速进行预测,提高车速预测的准确度;并将预测出的汽车短期车速运用到能量管理控制策略中,提高燃油经济性。

Description

一种基于前车与自车互动的短期车速工况实时预测方法
本申请要求于2019年11月25日提交中国专利局、申请号为201911168690.0、发明名称为“一种基于前车与自车互动的短期车速工况实时预测方法”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本发明涉及车辆的车速工况预测领域,特别是涉及一种基于前车与自车互动的短期车速工况实时预测方法。
背景技术
近年来,我国迅速增长的汽车需求带来了对石油消耗的急剧增长,同时也使我国所面临的能源安全问题更加突出。日益恶化的环境也促进了世界各国对节能减排的迫切需求。如何基于现有研究水平提高和完善混合动力系统能量管理算法的优化性能和理论体系,进一步实现基于优化的动力驱动控制算法的实时在线应用,是当前汽车发展中优化燃油消耗亟待解决的重要问题。
发明内容
本发明的目的是提供一种基于前车与自车互动的短期车速工况实时预测方法及系统,能够极大地提高了未来短期车速预测的精度,并将其运用到能量管理策略中,提高了燃油经济性。
为达到上述目的,本发明的技术方案为:
一种基于前车与自车互动的短期车速工况实时预测方法,包括:
获取自车历史车速信息、前车历史车速信息以及自车和前车之间的车距信息,并提取有效数据;所述有效数据包括自车运行时的有效实测数据车辆瞬时速度、前车运行时的有效实测数据车辆瞬时速度以及自车和前车运行时之间的瞬时车距;
根据所述自车历史车速信息、所述前车历史车速信息、所述自车和前车之间的车距信息以及所述有效数据构建未来车速预测模型;所述未来车速预测模型包括基于历史车速的预测模型以及基于历史车距的预测模型;
对构建的所述未来车速预测模型进行离线训练,确定训练后的未来车速预测模型;
根据所述训练后的未来车速预测模型在线预测自车的未来车速;所述未来车速为下一时刻的车速;
根据所述未来车速对所述训练后的未来车速预测模型自适应学习,确定自适应学习的未来车速预测模型;
根据所述自适应学习的未来车速预测模型预测自车的短期车速;所述短期车速为下一时段内的车速;
根据所述自车的短期车速计算扭矩需求;
基于动态规划算法,根据所述转矩需求计算最优转矩分配策略。
可选的,所述获取自车历史车速信息、前车历史车速信息以及自车和前车之间的车距信息,并提取有效数据,具体包括:
基于车载OBD或无线数据采集系统获取自车不同驾驶员在不同道路工况上行驶时的自车历史车速信息,存储于车速数据库中形成第一样本工况,并从所述第一样本工况中提取所述自车运行于各个样本工况时刻点t的有效实测数据车辆瞬时速度v 0(t);
运用车联网技术获取和所述自车对应的前车不同驾驶员在不同道路工况上行驶时的前车历史车速信息,存储于车速数据库中形成第二样本工况,并从所述第二样本工况中提取所述前车运行于各个样本工况时刻点t的有效实测数据车辆瞬时速度v 1(t);
运用毫米波雷达不断获取所述自车和所述前车之间的车距信息,存储于历史车距数据库中形成第三样本工况,并从所述第三样本工况中提取所述自车和所述前车运行于各个样本工况时刻点t的有效实测数据自车和前车运行时之间的瞬时车距S(t)。
可选的,所述根据所述自车历史车速信息、所述前车历史车速信息、所述自车和前车之间的车距信息以及所述有效数据构建未来车速预测模型,具体包括:
选取径向基函数人工神经网络作为非线性预测函数对自车未来车速进行预测,构建基于径向基函数人工神经网络的未来车速预测模型;所述径向基函数人工神经网络的未来车速预测模型包括输入层、隐藏层以及输出层;
所述输入层:
在基于历史车速信息构建模型时,选取驾驶员行为Q、所述自车在上一时间段H h秒内的历史车速和所述前车的上一时间段H s秒内的历史车速,作为所述未来车速预测模型的输入,定义为:
Figure PCTCN2020105463-appb-000001
或者,
在基于历史车距信息构建模型时,选取驾驶员行为Q、所述自车的上一时间段H r秒内的历史车速以及上一时间段H f秒内所述前车和所述自车之间的历史车距,作为所述未来车速预测模型的输入,定义为:
Figure PCTCN2020105463-appb-000002
其中,N in为预测模型输入,V k和V n分别是自车和前车当前时刻车速,V k-1和V n-1分别是自车和前车上一时刻的车速;S n是前车和自车当前时刻的车距,S n-1是前车和自车上一时刻的车距;H h,H s,H r和H f均为正整数;
所述隐藏层:
根据神经网络训练精度需求选定神经元数目O,隐藏层的激活函数为:
a 1=exp(-||n-c|| 2/2b 2),n=Wa 0+b;
其中,O为正整数,a 1和a 0分别是当前层和上一层的神经元输出,n是累计输出,c是神经元节点中心,b是神经元径向基函数扩散宽度,W是权重值;
所述输出层:
下一未来时间段H p秒内的车速作为输出,H p即为预测未来车速向量长度;假设f m和f n是神经网络预测的非线性函数;
在基于历史车速信息构建模型时,则有:
Figure PCTCN2020105463-appb-000003
在基于历史车距信息构建模型时,则有:
Figure PCTCN2020105463-appb-000004
可选的,所述对构建的所述未来车速预测模型进行离线训练,确定训练后的未来车速预测模型,具体包括:
构建输入矢量参数和输出矢量参数,并将输入参数矢量和输出参数矢量输入到基于径向基函数人工神经网络的未来车速预测模型模型中形成训练样本进行离线训练,建立稳定的径向基函数人工神经网络;
确定所述径向基函数人工神经网络为g-h-m的连接方式,即有g个输入,h个隐含层和m个输出;
选用自组织选取中线的多变量插值的径向基函数RBF神经网络学习方法,确定RBF神经网络中第j个输出表示为:
Figure PCTCN2020105463-appb-000005
式中,
Figure PCTCN2020105463-appb-000006
为第p个输入样本,p=1,2,…,P;P为样本总数,c i为网络隐含层节点的中心,i=1,2,…,h为隐含层的节点数,||x p-c i|| 2为欧式范数,σ i为基函数的宽度,ω ij为隐含层到输出层的连接权值, j=1,2,…,m为输出层的节点数,y j为与输入样本对应的神经网络的第j个输出节点的实际输出;
径向基函数人工神经网络离线训练步骤如下:
对权值ω ij赋初值为0到1之间的随机数,隐含层神经元的数目为h,初始网络误差E置0,最大误差ε设为一正的小数;
基于模糊K均值聚类算法确定基函数的中心c i及方差σ i,i=1,2,…,h;
采用梯度下降法调整网络隐含层到输出层的权值ω ij直到网络误差E<ε结束;其中网络误差采用均方误差来表示,表达式如下:
Figure PCTCN2020105463-appb-000007
式中,E表示网络误差,
Figure PCTCN2020105463-appb-000008
为对应于输入x p的实际输出,y(x p)表为对应于输入的期望输出,N为输入量x p的数量。
可选的,所述根据所述训练后的未来车速预测模型在线预测自车的未来车速,具体包括:
将所述训练后的未来车速预测模型嵌入到整车控制系统中;
基于所述整车控制系统,对所述自车下一时间段H p秒内的未来车速进行预测:
若所述训练后的未来车速预测模型为基于历史车速的模型,则运用自车上一时间段H h秒内的历史车速和前车上一时间段H s秒内的历史车速数据对自车下一时间段H p秒内的未来车速进行预测;
若所述训练后的未来车速预测模型为基于历史车距的模型,则运用自车上一时间段H r秒内的历史车速和前车和自车上一时间段H f秒内的历史车距数据对自车下一时间段H p秒内的未来车速进行预测;
在实车行驶过程中,预测未来车速的方式包括:
第一、基于车载OBD或无线数据采集系统不断获取自车上一时间段H h秒内的实时车速数据,基于车联网技术不断获取前车上一时间段H s秒 内的实时车速数据,并融合驾驶员风格形成神经网络输入矢量,实现自车未来车速预测;
第二、基于车载OBD或无线数据采集系统不断获取自车上一时间段H r秒内的实时车速数据,运用毫米波雷达不断获取前车和自车上一时间段H f秒内的实时车距数据,并融合驾驶员风格形成神经网络输入矢量,实现自车未来车速预测。
可选的,所述根据所述未来车速对所述训练后的未来车速预测模型自适应学习,确定自适应学习的未来车速预测模型,具体包括:
采集上一周期内自车和前车的车速数据或上一周期内前车和自车之间的车距数据,并更新样本车速或车距数据库;每个周期为一周、一月或一年;
利用更新的样本车速数据或车距数据,在车辆非行驶任务情况下,对所述训练后的未来车速预测模型进行再学习;
利用所述训练后的未来车速预测模型进行车速预测,确定自适应学习的未来车速预测模型。
可选的,所述扭矩需求的计算方式如下:
Figure PCTCN2020105463-appb-000009
其中,m为汽车质量,g为重力加速度,T out为输出扭矩,T break制动扭矩,R wheel为车轮半径,θ为道路坡度,C r为路面阻力系数,C d为风阻系数,A为车辆迎风面积。
可选的,所述基于动态规划算法,根据所述转矩需求计算最优转矩分配策略,具体包括:
获取划分阶段及选择阶段变量k、状态变量λ k、决策变量以及各级允许决策集合;
根据所述划分阶段及选择阶段变量k、所述状态变量λ k、所述决策变量以及所述各级允许决策集合确定状态转移方程:
λ k+1=T(λ k,v k);
其中,λ k+1为状态转移方程,T(λ k,v k)为状态变量为λ k、车速为v k时的转矩需求;
确定阶段目标函数的形式;所述目标函数具有可分性,并满足递推关系;
确定最优值函数满足的递推方程及端点条件,以r函数为代价函数:
r *k)=min[r(λ k,v k)+r *k+1)]
其中,r(λ k,v k)表示状态变量为λ k、车速为v k所对应的代价函数;r *k)表示状态变量为λ k时的最优代价函数;r *k+1)表示状态变量为λ k+1时的最优代价函数;
所述终端条件为:
r *k+1)=0
即求出r *k+1)为0时对应的k值;
逆序计算状态空间内最优价值函数,和最优解;所述最优解为在得到k值后,计算出所述最优价值函数对应的转矩需求T(λ k,v k);
根据所述最优价值函数和所述最优解,顺序计算给定初始状态下的最优转矩分配策略;所述最优转矩分配策略包括最优发动机转矩以及最优电机转矩控制变量。
本发明与现有技术相比的优点在于:本发明提供一种基于前车、自车互动模式,运用车联网技术,从人-车-环境系统的角度对实现汽车短期车速实时预测进行研究,通过神经网络在线学习,提出了一种综合考虑车辆状态参数、驾驶员驾驶风格以及前方车辆状态参数的车辆未来短期车速预测方法,提高了车速预测的准确度。并将提出的短期车速预测方法运用到燃油消耗控制策略中,获得更佳的燃油经济性。
说明书附图
下面结合附图对本发明作进一步说明:
图1为本发明的系统原理框图;
图2为实施例中的硬件设备示意图。
具体实施方式
面结合附图进一步详细描述本发明的技术方案,但本发明的保护范围不局限于以下所述。
如图1所示,一种基于前车与自车互动的短期车速工况实时预测方法,其特征在于:包括以下步骤:
S1.获取自车和前车历史车速、车距信息,并提取有效数据:
在本申请的实施例中,需要采用的设备包括但不限于,车载自动诊断系统(OBD),专用短程通讯(DSRC)系统中的车载单元(OBU)、路测单元(RSU),以及毫米波雷达。车载设备OBD可采集存储自车历史车速数据,路测单元(RSU)可获取前车历史车速数据,毫米波雷达可获取前后车车距信息,硬件设备如图2所示;步骤S1具体包括:
S101.基于车载OBD或无线数据采集系统获取自车不同驾驶员在不同道路工况上行驶时的历史车速信息,存储于车速数据库中形成第一样本工况;
S102.运用车联网技术(可选为“V2V”专用短程通讯设备)获取和自车对应的前车不同驾驶员在不同道路工况上行驶时的历史车速信息,存储于车速数据库中形成第二样本工况;
运用毫米波雷达不断获取前后车车距信息,存储于历史车距数据库中形成第三样本工况;
S103.从历史车速数据库第一样本工况中提取自车运行与各样本工况时刻点t的有效实测数据车辆瞬时速度v 0(t);
S104.从历史车速数据库第二样本工况中提取前车运行与各样本工况时刻点t的有效实测数据车辆瞬时速度v 1(t);
从历史车距数据库第三样本工况中提取前后车运行与各样本工况时刻点t的有效实测数据前后车瞬时车距S(t)。
S2.构建基于人工神经网络的未来车速预测模型:
人工神经网络可选为反向传播神经网络、层递归型神经网络、径向基函数神经网络等各类人工神经网络算法,在本申请的实施例中,选择径向基函数人工神经网络进行模型构建。
该步骤基于历史车速或历史车距信息构建人工神经网络的未来车速预测模型,具体包括以下步骤:
S201.选取径向基函数人工神经网络作为非线性预测函数对自车未来车速进行预测,构建基于径向基函数人工神经网络的未来车速预测模型;
S202.径向基函数人工神经网络由输入层、隐藏层、输出层三部分构成;
A1、在基于历史车速信息构建模型时,选取驾驶员行为Q、自车上一时间段H h秒内的历史车速和前车上一时间段H s秒内的历史车速,作为神经网络车速预测模型的输入,定义为:
Figure PCTCN2020105463-appb-000010
在基于历史车距信息构建模型时,选取驾驶员行为Q、自车上一时间段H r秒内的历史车速和前车、自车上一时间段H f秒内的历史车距,作为神经网络车速预测模型的输入,定义为:
Figure PCTCN2020105463-appb-000011
其中,N in为预测模型输入,V k和V n分别是自车和前车当前时刻车速,V k-1和V n-1分别是自车和前车上一时刻的车速;S n是前车和自车当前时刻的车距,S n-1是前车和自车上一时刻的车距;H h,H s,H r和H f均为正整数;
A2、根据神经网络训练精度需求选定神经元数目O,隐藏层的激活函数为:
a 1=exp(-||n-c|| 2/2b 2),n=Wa 0+b;
其中,O为正整数,a 1和a 0分别是当前层和上一层的神经元输出,n是累计输出,c是神经元节点中心,b是神经元径向基函数扩散宽度,W是权重值;
A3、下一未来时间段H p秒内的车速作为输出,H p即为预测未来车速向量长度,亦即神经网络模型输出向量的长度;假设f m和f n是神经网络预测的非线性函数;
在基于历史车速信息构建模型时,则有:
Figure PCTCN2020105463-appb-000012
在基于历史车距信息构建模型时,则有:
Figure PCTCN2020105463-appb-000013
S3.对构建的未来车速预测模型进行离线训练:
S301.按照步骤S2构建输入矢量参数和输出矢量参数,然后将输入参数矢量和输出参数矢量输入到径向基函数人工神经网络模型中形成训练样本进行离线训练,建立稳定的径向基函数人工神经网络结构;
在基于历史车速信息构建的模型中,输入矢量参数为:
Figure PCTCN2020105463-appb-000014
输出矢量参数为:
Figure PCTCN2020105463-appb-000015
在基于历史车距信息构建的模型中,输入矢量参数为:
Figure PCTCN2020105463-appb-000016
输出矢量参数为:
Figure PCTCN2020105463-appb-000017
S302.确定径向基函数人工神经网络为g-h-m的连接方式,即有g个 输入,h个隐含层和m个输出;
S303.选用自组织选取中线的RBF神经网络学习方法,核心是求解隐含层基函数中心、基函数的方差和隐含层单元到输出单元的权值,由此得RBF神经网络中第j个输出表示为:
Figure PCTCN2020105463-appb-000018
式中,
Figure PCTCN2020105463-appb-000019
为第p个输入样本,p=1,2,…,P;P为样本总数,ci为网络隐含层节点的中心,i=1,2,…,h为隐含层的节点数,||x p-c i|| 2为欧式范数,σ i为基函数的宽度,ω ij为隐含层到输出层的连接权值,j=1,2,…,m为输出层的节点数,y j为与输入样本对应的神经网络的第j个输出节点的实际输出;
径向基函数人工神经网络离线训练步骤如下:
对权值ω ij赋初值为0到1之间的随机数,隐含层神经元的数目为h,初始网络误差E置0,最大误差ε设为一正的小数;
基于模糊K均值聚类算法确定基函数的中心c i及方差σ i,i=1,2,…,h;
采用梯度下降法调整网络隐含层到输出层的权值ω ij直到网络误差E<ε,结束;其中网络误差采用均方误差来表示,表达式如下:
Figure PCTCN2020105463-appb-000020
式中,E表示网络误差,
Figure PCTCN2020105463-appb-000021
为对应于输入x p的实际输出,y(x p)表为对应于输入的期望输出,P为样本总数,N为输入x p的数量。
S4.在线预测自车的未来车速:
S401.将步骤S3得到的神经网络模型嵌入到整车控制系统中;
S402.对自车下一时间段H p秒内的未来车速进行预测:
若步骤S3中得到的模型为基于历史车速的模型,则运用自车上一时间段H h秒内的历史车速和前车上一时间段H s秒内的历史车速数据对自车下一时间段H p秒内的未来车速进行预测;
若步骤S3中得到的模型为基于历史车距的模型,则运用自车上一时间段H r秒内的历史车速和前车和自车上一时间段H f秒内的历史车距数据对自车下一时间段H p秒内的未来车速进行预测;
S403.在实车行驶过程中,预测未来车速的方式包括:
第一、基于车速采集系统不断获取自车上一时间段H h秒内的实时车速数据,基于车联网技术(可选为“V2V”专用短程通讯设备)不断获取前车上一时间段H s秒内的实时车速数据,并融合驾驶员风格形成神经网络输入矢量,实现自车未来车速预测;
第二、基于车速采集系统不断获取自车上一时间段H r秒内的实时车速数据,运用毫米波雷达不断获取前车和自车上一时间段H f秒内的实时车距数据,并融合驾驶员风格形成神经网络输入矢量,实现自车未来车速预测。
S5.实现车速预测神经网络的自适应学习:
S501.采集上一周期内自车和前车的车速数据或为前车和自车的车距数据,更新样本车速或车距数据库;每个周期为一周、一月或一年;
S502.利用更新的样本车速数据或车距数据,在车辆非行驶任务情况下,对神经网络预测模型进行再学习;
S503.利用步骤S502中得到的神经网络模型进行车速预测;
S504.收集新的样本数据后回到501,重复步骤S5的完整过程。
S6.根据预测的短期车速计算扭矩需求:
扭矩需求的计算方式如下:
Figure PCTCN2020105463-appb-000022
式中,m为汽车质量,T out为输出扭矩,T break制动扭矩,R wheel为车轮半径,θ为道路坡度,C r为路面阻力系数,C d为风阻系数,A为车辆迎风面积。
S7.根据转矩需求和动态规划算法计算最优转矩分配。
S701.划分阶段及选择阶段变量k;
S702.选择状态变量λ k
S703.选择决策变量及确定各级允许决策集合;
S704.写出状态转移方程,如下:
λ k+1=T(λ k,v k)
S705.确定阶段目标函数的形式,目标函数必须具有可分性,并满足递推关系;
S706.写出基本方程即最优值函数满足的递推方程及端点条件,以r函数为代价函数:
r *k)=min[r(λ k,v k)+r *k+1)]
终端条件:
r *n+1)=0
S707.逆序计算状态空间内最优价值函数,和对应的最优解;
S708.根据最优价值函数和最优解,顺序计算给定初始状态下的最优控制策略,即最优发动机转矩、电机转矩等控制变量。
综上,本发明基于前车、自车互动模式,运用车联网技术,从人-车-环境系统的角度对实现汽车短期车速实时预测进行研究,通过神经网络在线学习,提出了一种综合考虑车辆状态参数、驾驶员驾驶风格以及前方车辆状态参数的车辆未来短期车速预测方法,提高了车速预测的准确度。并将提出的短期车速预测方法运用到燃油消耗控制策略中,获得更佳的燃油经济性。
提供以上实施例仅仅是为了描述本发明的目的,而并非要限制本发明的范围。本发明的范围由所附权利要求限定。不脱离本发明的精神和原理而做出的各种等同替换和修改,均应涵盖在本发明的范围之内。

Claims (8)

  1. 一种基于前车与自车互动的短期车速工况实时预测方法,其特征在于,包括:
    获取自车历史车速信息、前车历史车速信息以及自车和前车之间的车距信息,并提取有效数据;所述有效数据包括自车运行时的有效实测数据车辆瞬时速度、前车运行时的有效实测数据车辆瞬时速度以及自车和前车运行时之间的瞬时车距;
    根据所述自车历史车速信息、所述前车历史车速信息、所述自车和前车之间的车距信息以及所述有效数据构建未来车速预测模型;所述未来车速预测模型包括基于历史车速的预测模型以及基于历史车距的预测模型;
    对构建的所述未来车速预测模型进行离线训练,确定训练后的未来车速预测模型;
    根据所述训练后的未来车速预测模型在线预测自车的未来车速;所述未来车速为下一时刻的车速;
    根据所述未来车速对所述训练后的未来车速预测模型自适应学习,确定自适应学习的未来车速预测模型;
    根据所述自适应学习的未来车速预测模型预测自车的短期车速;所述短期车速为下一时段内的车速;
    根据所述自车的短期车速计算扭矩需求;
    基于动态规划算法,根据所述转矩需求计算最优转矩分配策略。
  2. 根据权利要求1所述的基于前车与自车互动的短期车速工况实时预测方法,其特征在于,所述获取自车历史车速信息、前车历史车速信息以及自车和前车之间的车距信息,并提取有效数据,具体包括:
    基于车载OBD或无线数据采集系统获取自车不同驾驶员在不同道路工况上行驶时的自车历史车速信息,存储于车速数据库中形成第一样本工况,并从所述第一样本工况中提取所述自车运行于各个样本工况时刻点t的有效实测数据车辆瞬时速度v 0(t);
    运用车联网技术获取和所述自车对应的前车不同驾驶员在不同道路工况上行驶时的前车历史车速信息,存储于车速数据库中形成第二样本工 况,并从所述第二样本工况中提取所述前车运行于各个样本工况时刻点t的有效实测数据车辆瞬时速度v 1(t);
    运用毫米波雷达不断获取所述自车和所述前车之间的车距信息,存储于历史车距数据库中形成第三样本工况,并从所述第三样本工况中提取所述自车和所述前车运行于各个样本工况时刻点t的有效实测数据自车和前车运行时之间的瞬时车距S(t)。
  3. 根据权利要求2所述的基于前车与自车互动的短期车速工况实时预测方法,其特征在于,所述根据所述自车历史车速信息、所述前车历史车速信息、所述自车和前车之间的车距信息以及所述有效数据构建未来车速预测模型,具体包括:
    选取径向基函数人工神经网络作为非线性预测函数对自车未来车速进行预测,构建基于径向基函数人工神经网络的未来车速预测模型;所述径向基函数人工神经网络的未来车速预测模型包括输入层、隐藏层以及输出层;
    所述输入层:
    在基于历史车速信息构建模型时,选取驾驶员行为Q、所述自车在上一时间段H h秒内的历史车速和所述前车的上一时间段H s秒内的历史车速,作为所述未来车速预测模型的输入,定义为:
    Figure PCTCN2020105463-appb-100001
    或者,
    在基于历史车距信息构建模型时,选取驾驶员行为Q、所述自车的上一时间段H r秒内的历史车速以及上一时间段H f秒内所述前车和所述自车之间的历史车距,作为所述未来车速预测模型的输入,定义为:
    Figure PCTCN2020105463-appb-100002
    其中,N in为预测模型输入,V k和V n分别是自车和前车当前时刻车速,V k-1和V n-1分别是自车和前车上一时刻的车速;S n是前车和自车当前时刻的车距,S n-1是前车和自车上一时刻的车距;H h,H s,H r和H f均为正整数;
    所述隐藏层:
    根据神经网络训练精度需求选定神经元数目O,隐藏层的激活函数为:
    a 1=exp(-||n-c|| 2/2b 2),n=Wa 0+b;
    其中,O为正整数,a 1和a 0分别是当前层和上一层的神经元输出,n是累计输出,c是神经元节点中心,b是神经元径向基函数扩散宽度,W是权重值;
    所述输出层:
    下一未来时间段H p秒内的车速作为输出,H p即为预测未来车速向量长度;假设f m和f n是神经网络预测的非线性函数;
    在基于历史车速信息构建模型时,则有:
    Figure PCTCN2020105463-appb-100003
    在基于历史车距信息构建模型时,则有:
    Figure PCTCN2020105463-appb-100004
  4. 根据权利要求3所述的一种基于前车与自车互动的短期车速工况实时预测方法,其特征在于,所述对构建的所述未来车速预测模型进行离线训练,确定训练后的未来车速预测模型,具体包括:
    构建输入矢量参数和输出矢量参数,并将输入参数矢量和输出参数矢量输入到基于径向基函数人工神经网络的未来车速预测模型模型中形成训练样本进行离线训练,建立稳定的径向基函数人工神经网络;
    确定所述径向基函数人工神经网络为g-h-m的连接方式,即有g个输入,h个隐含层和m个输出;
    选用自组织选取中线的多变量插值的径向基函数RBF神经网络学习方法,确定RBF神经网络中第j个输出表示为:
    Figure PCTCN2020105463-appb-100005
    式中,
    Figure PCTCN2020105463-appb-100006
    为第p个输入样本,p=1,2,…,P;P为样本总数,c i为网络隐含层节点的中心,i=1,2,…,h为隐含层的节点数,||x p-c i|| 2为欧式范数,σ i为基函数的宽度,ω ij为隐含层到输出层的连接权值,j=1,2,…,m为输出层的节点数,y j为与输入样本对应的神经网络的第j个输出节点的实际输出;
    径向基函数人工神经网络离线训练步骤如下:
    对权值ω ij赋初值为0到1之间的随机数,隐含层神经元的数目为h,初始网络误差E置0,最大误差ε设为一正的小数;
    基于模糊K均值聚类算法确定基函数的中心c i及方差σ i,i=1,2,…,h;
    采用梯度下降法调整网络隐含层到输出层的权值ω ij直到网络误差E<ε结束;其中网络误差采用均方误差来表示,表达式如下:
    Figure PCTCN2020105463-appb-100007
    式中,E表示网络误差,
    Figure PCTCN2020105463-appb-100008
    为对应于输入x p的实际输出,y(x p)表为对应于输入的期望输出,N为输入x p的数量。
  5. 根据权利要求4所述的一种基于前车与自车互动的短期车速工况实时预测方法,其特征在于,所述根据所述训练后的未来车速预测模型在线预测自车的未来车速,具体包括:
    将所述训练后的未来车速预测模型嵌入到整车控制系统中;
    基于所述整车控制系统,对所述自车下一时间段H p秒内的未来车速进行预测:
    若所述训练后的未来车速预测模型为基于历史车速的模型,则运用自车上一时间段H h秒内的历史车速和前车上一时间段H s秒内的历史车速数据对自车下一时间段H p秒内的未来车速进行预测;
    若所述训练后的未来车速预测模型为基于历史车距的模型,则运用自车上一时间段H r秒内的历史车速和前车和自车上一时间段H f秒内的历史车距数据对自车下一时间段H p秒内的未来车速进行预测;
    在实车行驶过程中,预测未来车速的方式包括:
    第一、基于车载OBD或无线数据采集系统不断获取自车上一时间段H h秒内的实时车速数据,基于车联网技术不断获取前车上一时间段H s秒内的实时车速数据,并融合驾驶员风格形成神经网络输入矢量,实现自车未来车速预测;
    第二、基于车载OBD或无线数据采集系统不断获取自车上一时间段H r秒内的实时车速数据,运用毫米波雷达不断获取前车和自车上一时间段H f秒内的实时车距数据,并融合驾驶员风格形成神经网络输入矢量,实现自车未来车速预测。
  6. 根据权利要求5所述的一种基于前车与自车互动的短期车速工况实时预测方法,其特征在于,所述根据所述未来车速对所述训练后的未来车速预测模型自适应学习,确定自适应学习的未来车速预测模型,具体包括:
    采集上一周期内自车和前车的车速数据或上一周期内前车和自车之间的车距数据,并更新样本车速或车距数据库;每个周期为一周、一月或一年;
    利用更新的样本车速数据或车距数据,在车辆非行驶任务情况下,对所述训练后的未来车速预测模型进行再学习;
    利用所述训练后的未来车速预测模型进行车速预测,确定自适应学习的未来车速预测模型。
  7. 根据权利要求6所述的一种基于前车与自车互动的短期车速工况实时预测方法,其特征在于,所述扭矩需求的计算方式如下:
    Figure PCTCN2020105463-appb-100009
    其中,m为汽车质量,g为重力加速度,T out为输出扭矩,T break制动 扭矩,R wheel为车轮半径,θ为道路坡度,C r为路面阻力系数,C d为风阻系数,A为车辆迎风面积。
  8. 根据权利要求7所述的一种基于前车与自车互动的短期车速工况实时预测方法,其特征在于,所述基于动态规划算法,根据所述转矩需求计算最优转矩分配策略,具体包括:
    获取划分阶段及选择阶段变量k、状态变量λ k、决策变量以及各级允许决策集合;
    根据所述划分阶段及选择阶段变量k、所述状态变量λ k、所述决策变量以及所述各级允许决策集合确定状态转移方程:
    λ k+1=T(λ k,v k);
    其中,λ k+1为状态转移方程,T(λ k,v k)为状态变量为λ k、车速为v k时的转矩需求;
    确定阶段目标函数的形式;所述目标函数具有可分性,并满足递推关系;
    确定最优值函数满足的递推方程及端点条件,以r函数为代价函数:
    r *k)=min[r(λ k,v k)+r *k+1)]
    其中,r(λ k,v k)表示状态变量为λ k、车速为v k所对应的代价函数;r *k)表示状态变量为λ k时的最优代价函数;r *k+1)表示状态变量为λ k+1时的最优代价函数;
    所述终端条件为:
    r *k+1)=0
    即求出r *k+1)为0时对应的k值;
    逆序计算状态空间内最优价值函数,和最优解;所述最优解为在得到k值后,计算出所述最优价值函数对应的转矩需求T(λ k,v k);
    根据所述最优价值函数和所述最优解,顺序计算给定初始状态下的最优转矩分配策略;所述最优转矩分配策略包括最优发动机转矩以及最优电机转矩控制变量。
PCT/CN2020/105463 2019-11-25 2020-07-29 一种基于前车与自车互动的短期车速工况实时预测方法 WO2021103625A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911168690.0A CN111009134A (zh) 2019-11-25 2019-11-25 一种基于前车与自车互动的短期车速工况实时预测方法
CN201911168690.0 2019-11-25

Publications (1)

Publication Number Publication Date
WO2021103625A1 true WO2021103625A1 (zh) 2021-06-03

Family

ID=70113808

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/105463 WO2021103625A1 (zh) 2019-11-25 2020-07-29 一种基于前车与自车互动的短期车速工况实时预测方法

Country Status (2)

Country Link
CN (1) CN111009134A (zh)
WO (1) WO2021103625A1 (zh)

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113722837A (zh) * 2021-09-16 2021-11-30 吉林大学 根据不同工况的数据簇建立柴油车辆油耗与排放模型方法
CN113989534A (zh) * 2021-10-15 2022-01-28 厦门大学 一种基于多源运行数据的公交车站间载重识别方法和系统
CN114004993A (zh) * 2021-10-25 2022-02-01 厦门大学 基于lstm速度预测优化的ia-svm行驶工况识别方法及装置
CN114065402A (zh) * 2021-11-25 2022-02-18 华中科技大学 基于hssa-bpnn模型的车辆全态声品质预报方法
CN114228747A (zh) * 2022-01-25 2022-03-25 同济大学 基于贝叶斯网络的高速公路换道决策方法
CN114460936A (zh) * 2022-01-13 2022-05-10 华中科技大学 基于离线增量学习的自动驾驶汽车路径规划方法及系统
CN115880905A (zh) * 2022-11-29 2023-03-31 吉林大学 一种实时预测车辆通过交通信号灯路口时间的装置及方法
CN116331184A (zh) * 2023-03-31 2023-06-27 长安大学 基于机器视觉与雷达信息融合的车辆能量控制方法及系统
CN116522779A (zh) * 2023-04-29 2023-08-01 中国人民解放军63963部队 一种装甲车辆动力性与发动机运行map图匹配系统及方法
CN116702096A (zh) * 2023-08-04 2023-09-05 中汽研汽车检验中心(昆明)有限公司 车辆高原环境道路滑行阻力测算方法及装置
CN116821775A (zh) * 2023-08-29 2023-09-29 陕西重型汽车有限公司 一种基于机器学习的载重估计方法
CN117111620A (zh) * 2023-10-23 2023-11-24 山东省科学院海洋仪器仪表研究所 一种异构无人系统任务分配自主决策方法
CN117131955A (zh) * 2023-10-27 2023-11-28 北京航空航天大学 一种考虑多约束条件的短时车速预测方法
CN117746066A (zh) * 2024-02-20 2024-03-22 贵州博睿科讯科技发展有限公司 一种扩散模型引导的高速车辆检测集成学习方法和装置
CN115880905B (zh) * 2022-11-29 2024-05-17 吉林大学 一种实时预测车辆通过交通信号灯路口时间的装置及方法

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111009134A (zh) * 2019-11-25 2020-04-14 北京理工大学 一种基于前车与自车互动的短期车速工况实时预测方法
CN111791887B (zh) * 2020-07-03 2021-06-04 北京理工大学 一种基于分层式车速规划的车辆节能驾驶方法
CN112785845B (zh) * 2020-12-30 2022-11-01 桂林电子科技大学 一种基于K-means聚类与RBF神经网络的车速预测方法
CN112937547B (zh) * 2021-01-28 2023-03-24 北京理工大学 一种基于全局工况的插电式混合动力公交车能量管理方法
CN113267345A (zh) * 2021-04-23 2021-08-17 联合汽车电子有限公司 车辆前方未知路段阻力预测方法、存储介质、控制器和系统
CN113212414B (zh) * 2021-06-16 2023-03-24 北京理工大学 一种串联式机电复合传动系统的动力储备预测控制方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005091112A (ja) * 2003-09-16 2005-04-07 Equos Research Co Ltd 走行速度パターン推定装置、及びハイブリッド車両の駆動制御装置
CN103914985A (zh) * 2014-04-25 2014-07-09 大连理工大学 一种混合动力客车未来车速轨迹预测方法
CN109131350A (zh) * 2018-08-23 2019-01-04 北京理工大学 一种混合动力汽车能量管理方法及系统
CN111009134A (zh) * 2019-11-25 2020-04-14 北京理工大学 一种基于前车与自车互动的短期车速工况实时预测方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4786553B2 (ja) * 2007-01-04 2011-10-05 トヨタ自動車株式会社 回転数の予測装置、予測方法、その方法を実現させるプログラムおよびそのプログラムを記録した記録媒体
EP3006698B1 (en) * 2013-05-31 2018-08-22 Hitachi Automotive Systems, Ltd. Vehicle control apparatus and vehicle control method
KR101994302B1 (ko) * 2016-11-02 2019-09-30 현대자동차주식회사 하이브리드 자동차 및 그를 위한 변속 제어 방법
CN109649389B (zh) * 2018-11-05 2020-06-26 江苏大学 一种加速度可调节的自适应巡航控制系统及控制方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005091112A (ja) * 2003-09-16 2005-04-07 Equos Research Co Ltd 走行速度パターン推定装置、及びハイブリッド車両の駆動制御装置
CN103914985A (zh) * 2014-04-25 2014-07-09 大连理工大学 一种混合动力客车未来车速轨迹预测方法
CN109131350A (zh) * 2018-08-23 2019-01-04 北京理工大学 一种混合动力汽车能量管理方法及系统
CN111009134A (zh) * 2019-11-25 2020-04-14 北京理工大学 一种基于前车与自车互动的短期车速工况实时预测方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
HUANG GUANGQIAN, GUANGYU YONG TIAN: "Research on Predicting and Tracking Algorithm of SOC Trajectory for PHEV with Traffic Information Considered", JOURNAL OF CHONGQING UNIVERSITY OF TECHNOLOGY(NATURAL SCIENCE), vol. 32, no. 8, 15 August 2018 (2018-08-15), pages 1 - 7, XP055814627, ISSN: 1674-8425, DOI: 10.3969/j.issn.1674-8425(z).2018.08.001 *

Cited By (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113722837B (zh) * 2021-09-16 2023-09-29 吉林大学 根据不同工况的数据簇建立柴油车辆油耗与排放模型方法
CN113722837A (zh) * 2021-09-16 2021-11-30 吉林大学 根据不同工况的数据簇建立柴油车辆油耗与排放模型方法
CN113989534A (zh) * 2021-10-15 2022-01-28 厦门大学 一种基于多源运行数据的公交车站间载重识别方法和系统
CN114004993A (zh) * 2021-10-25 2022-02-01 厦门大学 基于lstm速度预测优化的ia-svm行驶工况识别方法及装置
CN114065402A (zh) * 2021-11-25 2022-02-18 华中科技大学 基于hssa-bpnn模型的车辆全态声品质预报方法
CN114065402B (zh) * 2021-11-25 2024-04-19 华中科技大学 基于hssa-bpnn模型的车辆全态声品质预报方法
CN114460936A (zh) * 2022-01-13 2022-05-10 华中科技大学 基于离线增量学习的自动驾驶汽车路径规划方法及系统
CN114460936B (zh) * 2022-01-13 2024-04-02 华中科技大学 基于离线增量学习的自动驾驶汽车路径规划方法及系统
CN114228747A (zh) * 2022-01-25 2022-03-25 同济大学 基于贝叶斯网络的高速公路换道决策方法
CN114228747B (zh) * 2022-01-25 2023-04-25 同济大学 基于贝叶斯网络的高速公路换道决策方法
CN115880905A (zh) * 2022-11-29 2023-03-31 吉林大学 一种实时预测车辆通过交通信号灯路口时间的装置及方法
CN115880905B (zh) * 2022-11-29 2024-05-17 吉林大学 一种实时预测车辆通过交通信号灯路口时间的装置及方法
CN116331184B (zh) * 2023-03-31 2024-04-02 长安大学 基于机器视觉与雷达信息融合的车辆能量控制方法及系统
CN116331184A (zh) * 2023-03-31 2023-06-27 长安大学 基于机器视觉与雷达信息融合的车辆能量控制方法及系统
CN116522779A (zh) * 2023-04-29 2023-08-01 中国人民解放军63963部队 一种装甲车辆动力性与发动机运行map图匹配系统及方法
CN116522779B (zh) * 2023-04-29 2023-10-20 中国人民解放军63963部队 一种装甲车辆动力性与发动机运行map图匹配系统及方法
CN116702096B (zh) * 2023-08-04 2023-10-03 中汽研汽车检验中心(昆明)有限公司 车辆高原环境道路滑行阻力测算方法及装置
CN116702096A (zh) * 2023-08-04 2023-09-05 中汽研汽车检验中心(昆明)有限公司 车辆高原环境道路滑行阻力测算方法及装置
CN116821775A (zh) * 2023-08-29 2023-09-29 陕西重型汽车有限公司 一种基于机器学习的载重估计方法
CN117111620A (zh) * 2023-10-23 2023-11-24 山东省科学院海洋仪器仪表研究所 一种异构无人系统任务分配自主决策方法
CN117111620B (zh) * 2023-10-23 2024-03-29 山东省科学院海洋仪器仪表研究所 一种异构无人系统任务分配自主决策方法
CN117131955A (zh) * 2023-10-27 2023-11-28 北京航空航天大学 一种考虑多约束条件的短时车速预测方法
CN117131955B (zh) * 2023-10-27 2024-01-16 北京航空航天大学 一种考虑多约束条件的短时车速预测方法
CN117746066A (zh) * 2024-02-20 2024-03-22 贵州博睿科讯科技发展有限公司 一种扩散模型引导的高速车辆检测集成学习方法和装置
CN117746066B (zh) * 2024-02-20 2024-05-07 贵州博睿科讯科技发展有限公司 一种扩散模型引导的高速车辆检测集成学习方法和装置

Also Published As

Publication number Publication date
CN111009134A (zh) 2020-04-14

Similar Documents

Publication Publication Date Title
WO2021103625A1 (zh) 一种基于前车与自车互动的短期车速工况实时预测方法
CN111267831B (zh) 一种混合动力车辆智能变时域模型预测能量管理方法
Wu et al. Continuous reinforcement learning of energy management with deep Q network for a power split hybrid electric bus
Huang et al. Model predictive control power management strategies for HEVs: A review
Yuan et al. Intelligent energy management strategy based on hierarchical approximate global optimization for plug-in fuel cell hybrid electric vehicles
CN107688343B (zh) 一种混合动力车辆的能量控制方法
CN109895760B (zh) 基于soc参考轨迹的插电式混合动力汽车能量管理方法
WO2021114742A1 (zh) 一种混合动力电动汽车综合预测能量管理方法
Sun et al. Velocity predictors for predictive energy management in hybrid electric vehicles
CN103914985B (zh) 一种混合动力客车未来车速轨迹预测方法
Lin et al. An ensemble learning velocity prediction-based energy management strategy for a plug-in hybrid electric vehicle considering driving pattern adaptive reference SOC
CN112668799A (zh) 基于行驶大数据的phev的智能能量管理方法和存储介质
CN113525396B (zh) 一种融合深度强化学习的混动汽车分层预测能量管理方法
CN116070783B (zh) 一种混动传动系统在通勤路段下的学习型能量管理方法
Li et al. A Kriging-based bi-objective constrained optimization method for fuel economy of hydrogen fuel cell vehicle
CN114312845A (zh) 基于地图数据的深度强化学习型混合动力汽车控制方法
Montazeri-Gh et al. Driving condition recognition for genetic-fuzzy HEV control
CN108082188A (zh) 车辆控制单元(vcu)及其操作方法
CN115107733A (zh) 一种混合动力汽车的能量管理方法及系统
CN115805840A (zh) 一种增程式电动装载机能耗控制方法及系统
CN115257695A (zh) 一种混合动力汽车能量管理控制系统及其方法
CN115534929A (zh) 基于多元信息融合的插电式混合动力汽车能量管理方法
Xia et al. A predictive energy management strategy for multi-mode plug-in hybrid electric vehicle based on long short-term memory neural network
Gan et al. Intelligent learning algorithm and intelligent transportation-based energy management strategies for hybrid electric vehicles: a review
CN113135113B (zh) 一种全局soc规划方法及装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20893711

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20893711

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 29/03/2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20893711

Country of ref document: EP

Kind code of ref document: A1