WO2022237212A1 - 基于复杂网络的自动驾驶汽车复杂环境模型、认知***及认知方法 - Google Patents
基于复杂网络的自动驾驶汽车复杂环境模型、认知***及认知方法 Download PDFInfo
- Publication number
- WO2022237212A1 WO2022237212A1 PCT/CN2022/070671 CN2022070671W WO2022237212A1 WO 2022237212 A1 WO2022237212 A1 WO 2022237212A1 CN 2022070671 W CN2022070671 W CN 2022070671W WO 2022237212 A1 WO2022237212 A1 WO 2022237212A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- complex
- nodes
- node
- network
- driving
- Prior art date
Links
- 230000019771 cognition Effects 0.000 title claims abstract description 48
- 230000001149 cognitive effect Effects 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000012546 transfer Methods 0.000 claims abstract description 14
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 10
- 239000011159 matrix material Substances 0.000 claims description 28
- 230000008859 change Effects 0.000 claims description 26
- 230000008878 coupling Effects 0.000 claims description 26
- 238000010168 coupling process Methods 0.000 claims description 26
- 238000005859 coupling reaction Methods 0.000 claims description 26
- 238000007637 random forest analysis Methods 0.000 claims description 24
- 230000006870 function Effects 0.000 claims description 13
- 230000004069 differentiation Effects 0.000 claims description 10
- 230000007704 transition Effects 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 7
- 230000001133 acceleration Effects 0.000 claims description 6
- 238000005054 agglomeration Methods 0.000 claims description 6
- 230000002776 aggregation Effects 0.000 claims description 6
- 238000013145 classification model Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 claims description 6
- 238000003066 decision tree Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims 1
- 238000000691 measurement method Methods 0.000 abstract description 3
- SAZUGELZHZOXHB-UHFFFAOYSA-N acecarbromal Chemical compound CCC(Br)(CC)C(=O)NC(=O)NC(C)=O SAZUGELZHZOXHB-UHFFFAOYSA-N 0.000 description 4
- 230000007613 environmental effect Effects 0.000 description 4
- 230000008447 perception Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000005079 cognition system Anatomy 0.000 description 1
- 231100000870 cognitive problem Toxicity 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000005312 nonlinear dynamic Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
- B60W40/08—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models related to drivers or passengers
- B60W40/09—Driving style or behaviour
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units
- B60W30/18—Propelling the vehicle
- B60W30/182—Selecting between different operative modes, e.g. comfort and performance modes
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
- B60W40/10—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models related to vehicle motion
- B60W40/107—Longitudinal acceleration
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W50/00—Details of control systems for road vehicle drive control not related to the control of a particular sub-unit, e.g. process diagnostic or vehicle driver interfaces
- B60W50/08—Interaction between the driver and the control system
- B60W50/082—Selecting or switching between different modes of propelling
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W2520/00—Input parameters relating to overall vehicle dynamics
- B60W2520/10—Longitudinal speed
- B60W2520/105—Longitudinal acceleration
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W2540/00—Input parameters relating to occupants
- B60W2540/30—Driving style
Definitions
- the invention relates to the technical field of self-driving cars, in particular to a complex network-based self-driving car complex environment model, a cognitive system and a cognitive method.
- a complex network is a highly complex network, an abstraction of a complex system, and generally has some or all of the properties of self-organization, self-similarity, attractor, small world, and scale-free.
- the characteristics of a complex network are complexity, which is specifically manifested in: large network scale, complex connection structure, node complexity (such as: node dynamic complexity and node diversity), complex spatiotemporal evolution process of the network, sparseness of network connections, multiple A heavy complexity fusion etc.
- Complexity research methods for complex networks such as node complexity, connection structure complexity, and network spatiotemporal evolution process complexity, have become important tools for modeling and researching complex systems.
- An autonomous vehicle is a comprehensive system that integrates environmental perception, planning and decision-making, and control execution. Due to the rapid development of sensor technologies such as lidar, millimeter-wave radar, and camera, environmental perception methods have been studied in depth and great progress has been made. At present, the establishment of the relationship between the individual type, location, movement and other low-level perception information of the environment and the individual behavior style, hierarchical local environment, and global environmental cognition supports the transition from environmental perception to individual cognition, local cognition to traffic comprehensive situation The development of global cognition has become an important prerequisite for ensuring the safety of autonomous decision-making and motion planning of autonomous vehicles. However, the environment faced by self-driving cars is a complex system.
- the present invention provides a complex network-based self-driving car complex environment model, a cognitive system and a cognitive method.
- the driving style is identified based on the driving characteristic parameters used to represent the aggressiveness of driving manipulation and the mode transfer preference; secondly, based on the group behavior characteristics of the moving subjects in the environment, on the basis of the driving style identification, based on the complex network,
- a time-varying complex dynamic network is established as the complex environment model of the autonomous vehicle;
- the nodes in the complex environment model are parametrically expressed to realize the node differential cognition of the complex environment,
- the agglomerative algorithm is used to stratify the nodes in the complex environment model to realize the hierarchical cognition of the complex environment, establish the disorder degree measurement method of the complex environment model, and realize the global risk situation cognition of the complex environment.
- the cognition system of the self-driving car based on the complex network of the present invention includes: a driving style recognition module, a complex environment model module, a node differentiation cognition module, a hierarchical cognition module, and a global risk situation cognition module.
- the driving style recognition module constructs a driving style feature matrix C J on the basis of extracting driving feature parameters, inputs the driving style feature matrix C J into a random forest classifier R f , and outputs the driving style through the random forest classifier R f Category K drive .
- the driving characteristic parameters include longitudinal driving characteristic parameters, lateral driving characteristic parameters and mode transfer characteristic parameters.
- the longitudinal driving characteristic parameters refer to the longitudinal acceleration a + and the heel-relaxation time d time within the limited time window
- the lateral driving characteristic parameters refer to the lateral acceleration RMS(a - ), lateral
- the driving style characteristic matrix C J refers to a three-dimensional six-degree-of-freedom characteristic matrix composed of longitudinal driving characteristic parameters, lateral driving characteristic parameters and mode transfer characteristic parameters:
- the random forest classifier R f is generated through the following steps: the original training set composed of driving style data is randomly sampled with replacement, m training sets are generated, n features are selected for each training set, and m decision-making Tree classification model, for each classification model, select the best sample features according to the information gain ratio to split until all training samples belong to the same class, and finally form all the generated decision tree classification models into a random forest, and output driving by voting method Style category K drive .
- the driving style category K drive includes three categories: aggressive, peaceful, and conservative:
- the complex environment model module is to describe the random, dynamic and nonlinear evolution law of the complex environment of the self-driving car, based on the complex network theory, with the moving subject as the node, constructing a time-varying complex dynamic network G as the complex environment model:
- G is a time-varying complex dynamic network
- V is a set of nodes in a time-varying complex dynamic network G
- B is a set of edges in a time-varying complex dynamic network G, representing the connection between nodes
- X is a time-varying complex dynamic network
- P is the strength function of the edge in the complex dynamic network G, which represents the coupling relationship between nodes
- ⁇ is the area function of the time-varying complex dynamic network G, which represents the dynamic constraints on the time-varying complex dynamic network G.
- the time-varying complex dynamic network G is equivalent to a continuous-time dynamic system with N nodes, and the state variable of the i-th node is set to x i , then the dynamic equation of the i-th node is:
- f(x i ) is the independent function of the state variable of the i-th node
- ⁇ >0 is the strength coefficient of the common connection relationship
- p ij (t) is the coupling coefficient between the i-th node and the j-th node
- H(x j ) is an inline function between nodes, which is a function of driving style and node distance.
- X is the state vector of the nodes in the time-varying complex dynamic network G
- F(X) is the dynamic equation vector of the nodes in the time-varying complex dynamic network G
- P(t) is the coupling between nodes in the time-varying complex dynamic network G matrix
- H(X) is the inline vector of nodes in the time-varying complex dynamic network G.
- the node differentiation cognition module expresses the difference of network nodes with four parameters including the quantity g i , degree k i , point weight s i and importance I(i) of the nodes in the complex environment model, and uses Normal distribution plots are differentiated across all nodes.
- the quantity g i of the nodes is represented by the structure size of the i-th node.
- the degree ki of the node is represented by the number of nodes directly connected to the i-th node.
- the point weight s i of the node represents the sum of edge weights of all adjacent edges of the i-th node.
- p ij (t) is the coupling coefficient between nodes
- K(i) is the degree centrality factor of the i-th node:
- ⁇ k> ⁇ k i /N, which represents the average degree of the module; Indicates the average unit weight of the module.
- the hierarchical cognition module adopts the agglomeration algorithm to divide the nodes in the complex environment model hierarchically, so as to realize the hierarchical and stepwise cognition of the complex environment of the self-driving car.
- the operation steps are as follows:
- the first step is to take the self-driving car as the central node, and the nodes that have a coupling relationship with the central node and the central node form the inner module;
- the second step is to sort the non-central nodes of the inner module by importance, and find the point with the largest coupling coefficient in turn to form the middle module;
- the third step is to sort the importance of the nodes of the middle layer module, and then find the point with the largest coupling coefficient to form the outer layer module;
- the global risk situation cognition module is based on the basic idea of entropy theory, uses system entropy and entropy change to measure the degree of disorder of the complex environment model, describes the overall risk and change situation, and realizes the state cognition of the global commonality.
- V n is the number of nodes in the complex environment model
- ⁇ is the network area of the complex environment model
- D(P) represents the variance of the coupling coefficient
- D(U) is the variance of the node speed in the complex environment model.
- d means to calculate the differential of the corresponding variable, indicating its changing trend.
- the cognitive method of the self-driving car proposed by the present invention includes the following steps:
- Step 1) Extract longitudinal driving characteristic parameters, lateral driving characteristic parameters and mode transfer characteristic parameters, construct driving style characteristic matrix C J , generate random forest classifier R f , input driving style characteristic matrix C J into random forest classifier R f , The output driving style category K drive of the random forest classifier R f recognizes the driving style as aggressive, peaceful and conservative;
- Step 2) Construct a time-varying complex dynamic network G as a complex environment model, which is used to describe the overall correlation characteristics of the complex environment, further establish the node dynamic equation in the complex environment model, and then combine the characteristics of all nodes in the time-varying complex dynamic network G to form The dynamic equation vector F(X), the coupling matrix P(t) between nodes in the time-varying complex dynamic network G and the inline vector H(X) of the nodes, establish the node system dynamic equation of the time-varying complex dynamic network G, using to describe the dynamic characteristics of complex environments;
- Step 3) Construct four parameters of the node quantity g i , degree k i , point weight s i and importance I(i) in the complex environment model, and use the normal distribution graph to conduct differential analysis on nodes to realize node differential recognition Know;
- Step 4) Use the agglomeration algorithm to divide the nodes in the complex environment model into layers, so as to realize the hierarchical and step-by-step cognition of the complex environment of the self-driving car;
- Step 5 According to the basic idea of entropy theory, use system entropy and entropy change to measure the degree of disorder of the complex environment model, describe the overall risk and change situation, and realize the state cognition of the global commonality.
- the present invention first aims at the complexity of individual driving behavior cognition, and performs driving style recognition according to the driving characteristic parameters indicating the aggressiveness of driving manipulation and mode transfer preference; secondly, according to the complex Group behavior characteristics of moving subjects in the environment, on the basis of driving style recognition, based on complex networks, with moving subjects as nodes and roads as constraints, construct a time-varying complex dynamic network G as a complex environment model for autonomous vehicles; finally, the The nodes in the complex environment model are expressed parametrically to realize the differential cognition of the nodes in the complex environment, and the nodes in the complex environment model are layered by using the agglomeration algorithm, so as to realize the hierarchical cognition of the complex environment and establish the complex environment model
- the measurement method of the degree of disorder can realize the global risk situation cognition of the complex environment, so as to establish the complex environment model, cognitive method and device of the self-driving car based on the complex network, and lay a solid foundation for the safe driving
- the present invention establishes a driving style recognition method. On the basis of extracting the driving characteristic parameters, the driving style characteristic matrix C J is constructed, and the driving style characteristic matrix C J is input into the random forest classifier R f , and the random forest classifier R f outputs Driving style category K drive to realize driving style recognition;
- the present invention is based on the complex network theory, takes the moving subject as the node, constructs a time-varying complex dynamic network G as a complex environment model, describes the random, dynamic, and nonlinear evolution law of the complex environment of the self-driving car, and also establishes a time-varying complex dynamic network G.
- the dynamic equation of the node system of the dynamic network G which describes the dynamic characteristics of the complex environment;
- the present invention constructs four parameters of nodes in the complex environment model, namely, the quantity g i , degree k i , point weight s i , and importance I(i), and uses a normal distribution diagram to perform differential analysis on the nodes, so as to realize automatic driving Differentiated cognition of nodes in the complex environment of automobiles;
- the present invention adopts the agglomeration algorithm to divide the nodes in the complex environment model hierarchically, so as to realize the hierarchical and stepwise cognition of the complex environment of the self-driving car;
- the present invention constructs the system entropy and entropy change of the complex environment model of the self-driving car to measure the degree of disorder of the complex environment model, describes the overall risk and change situation, and realizes the state recognition of the global commonality of the complex environment of the self-driving car.
- Figure 1 is a flow chart of the structure of the driving style recognition module.
- Fig. 2 Flow chart of the module structure of the complex environment model of the self-driving car.
- Figure 3 is a structural diagram of the node differentiation cognitive module.
- Figure 4 is a flow chart of the hierarchical cognitive module structure.
- Figure 5 Structural diagram of the global risk situation awareness module.
- Figure 6 is a schematic structural diagram of a self-driving car cognitive system based on a complex network.
- the longitudinal driving characteristic parameters refer to the longitudinal acceleration a + , Heel-relaxation time d time
- the lateral driving characteristic parameters refer to the lateral acceleration root mean square RMS(a - ) and the yaw rate standard deviation SD(r) within a limited time window
- the mode transfer characteristic parameters have a limited time window
- the left lane-changing state transition probability P(l c ) and the right lane-changing state transition probability P(r c ) then, construct the driving style characteristic matrix C J
- the driving style characteristic matrix C J refers to the A three-dimensional six-degree-of-freedom feature matrix composed of feature parameters, lateral driving feature parameters, and mode transfer feature parameters; then, input the driving style feature matrix C J into the random forest classifier R f , and output the driving style category K drive
- the driving style category K drive includes three types: aggressive, peaceful, and
- G (V, B, X, P, ⁇ )
- the time-varying complex The dynamic network G is equivalent to a continuous-time dynamic system with N nodes, and the dynamic equation of the nodes is established:
- the dynamic equation of the node system is established: Finally, the node system dynamic equation is input into the complex environment model to describe the dynamic characteristics of the complex environment.
- the node differentiated cognitive module structure uses four parameters of the node quantity g i , degree ki , point weight s i and importance I(i) in the complex environment model to describe the network node
- the differences of all nodes are analyzed using the normal distribution graph to realize the differential cognition of the nodes.
- the hierarchical cognition module structure process adopts the agglomerative algorithm to divide the nodes in the complex environment model hierarchically, and divides the nodes in the complex environment model in turn and forms the inner module, the middle module, and the outer module respectively.
- the layer module and the edge layer module realize the hierarchical cognition of the complex environment.
- the cognitive system of an autonomous vehicle based on a complex network includes a driving style recognition module, a complex environment model module, a node differentiation cognitive module, a hierarchical cognitive module, and a global risk situation cognitive module.
- the driving style identification module inputs the identified node driving style into the complex environment model module, which is used to construct the inter-node interlink function H(x j ); the node differentiation cognitive module, hierarchical cognitive module, global risk
- the situational cognition module receives the data of V, B, X, P, and ⁇ parameters in the complex environment model module, and realizes node differential cognition, hierarchical cognition and global risk situation cognition respectively.
- a cognitive method for a self-driving car based on a complex network includes the following steps:
- Step 1) Extract longitudinal driving characteristic parameters, lateral driving characteristic parameters and mode transfer characteristic parameters, construct driving style characteristic matrix C J , generate random forest classifier R f , input driving style characteristic matrix C J into random forest classifier R f , The output driving style category K drive of the random forest classifier R f identifies the driving style as aggressive, peaceful and conservative.
- the specific steps are:
- Step 2 Construct a time-varying complex dynamic network G as a complex environment model, which is used to describe the overall correlation characteristics of the complex environment, further establish the node dynamic equation in the complex environment model, and then combine the characteristics of all nodes in the time-varying complex dynamic network G to form The dynamic equation vector F(X), the coupling matrix P(t) between nodes in the time-varying complex dynamic network G and the inline vector H(X) of the nodes, establish the node system dynamic equation of the time-varying complex dynamic network G, using To describe the dynamic characteristics of complex environments, the specific steps are:
- Step 3) Construct four parameters of the node quantity g i , degree k i , point weight s i and importance I(i) in the complex environment model, and use the normal distribution diagram to perform differential analysis on all nodes to realize node differentiation cognition, the specific steps are:
- Step 4) Use the agglomerative algorithm to divide the nodes in the complex environment model hierarchically, and realize the hierarchical and step-by-step cognition of the complex environment of the self-driving car.
- the specific steps are:
- Step 5 According to the basic idea of entropy theory, use system entropy and entropy change to measure the degree of disorder of the complex environment model, describe the overall risk and change situation, and realize the state cognition of the global commonality.
- the specific steps are as follows:
- Specific embodiments of the present invention use Python to write the driving style recognition module, construct the driving style feature matrix C J based on the Scikit-learn third-party machine learning library, generate a random forest classifier R f , and realize the driving style recognition; use MATLAB/Simulink to write The mathematical model constitutes a complex environment model module; use Python to write the node differentiation cognitive module, hierarchical cognitive module, and global risk situation cognitive module, and realize the differentiation, hierarchy, and global risk of the complex environment of autonomous vehicles in the PyTorch framework Situational awareness method; write MATLAB, Scikit-learn and PyTorch interfaces based on Ubuntu system, install and configure them in industrial control computers, and realize complex environment models, cognitive methods and devices for autonomous vehicles based on complex networks.
Landscapes
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Transportation (AREA)
- Mechanical Engineering (AREA)
- Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Human Computer Interaction (AREA)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
- Control Of Driving Devices And Active Controlling Of Vehicle (AREA)
- Feedback Control In General (AREA)
Abstract
Description
Claims (10)
- 基于复杂网络的自动驾驶汽车复杂环境模型,其特征在于,以运动主体为节点,构造时变复杂动态网络G作为复杂环境模型:G=(V,B,X,P,Θ) (3)其中,G是时变复杂动态网络,V是时变复杂动态网络G中节点集合,B是时变复杂动态网络G中边的集合,表示节点之间的连线,X是时变复杂动态网络G中节点的状态向量,P为复杂动态网络G中边的强度函数,表示节点间的耦合关系,Θ为时变复杂动态网络G的区域函数,表示对时变复杂动态网络G的动态约束;将时变复杂动态网络G等效为具有N个节点的连续时间动态***,设第i节点的状态变量为x i,则第i节点的动力学方程为:其中,f(x i)为第i节点状态变量的自变函数,ξ>0为共同连接关系强度系数,p ij(t)为第i节点和第j节点之间的耦合系数,H(x j)为节点间的内联函数,是驾驶风格和节点距离的函数;记X=[x 1,x 2,…,x N] T,F(X)=[f(x 1),f(x 2),…,f(x N)] T,P(t)=[(p ij(t))]∈R N×N,H(X)=[H(x 1),H(x 2),…,H(x N)] T,则时变复杂动态网络G的节点***动力学方程为:其中,X为时变复杂动态网络G中节点的状态向量,F(X)为时变复杂动态网络G中节点的动态方程向量,P(t)为时变复杂动态网络G中节点间的耦合矩阵,H(X)为时变复杂动态网络G中节点的内联向量;所述复杂环境模型中,随着节点的运动和环境的变化,节点的位置和状态处于动态变化中,有节点汇入和流出网络,节点间的耦合关系和网络区域函数也随之变化,复杂网络***随时间不断地演化发展。
- 基于复杂网络的自动驾驶汽车的认知***,其特征在于,包括:驾驶风格识别模块,复杂环境模型模块,节点差异化认知模块,层次化认知模块,全局风险态势认知模块;所述驾驶风格识别模块,是在提取驾驶特征参数的基础上,构造驾驶风格特征矩阵C J,将驾驶风格特征矩阵C J输入随机森林分类器R f,通过随机森林分类器R f输出驾驶风格类别K drive;所述复杂环境模型模块为权利要求1所述的复杂环境模型;所述节点差异化认知模块,利用复杂环境模型中节点的量g i、度k i、点权s i和重要度I(i)共四个参数来表述了网络节点的差异性,并用正态分布图对所有节点进行差异化分析;所述层次化认知模块,是采用凝聚算法对复杂环境模型中的节点进行层次划分,实现对自动驾驶汽车复杂环境的层次化、阶梯性认知;所述全局风险态势认知模块,利用***熵和熵变对复杂环境模型的无序程度进行度量,描述整体风险及变化态势,实现对全局共性的状态认知。
- 根据权利要求2所述的自动驾驶汽车的认知***,其特征在于,所述驾驶特征参数,包括纵向驾驶特征参数、横向驾驶特征参数和模式转移特征参数。所述纵向驾驶特征参数是指有限时窗内的纵向加速度a +、跟弛时距d time,所述横向驾驶特征参数是指有限时窗内的横向加速度均方根RMS(a -)、横摆角速度标准差SD(r),所述模式转移特征参数有限时窗内的左换道状态转移概率P(l c)和右换道状态转移概率P(r c)。
- 根据权利要求2所述的自动驾驶汽车的认知***,其特征在于,所述随机森林分类器R f通过如下步骤生成:对驾驶风格数据组成的原始训练集进行有放回随机抽样,生成m个训练集,对每个训练集选择n个特征,分别训练m个决策树分类模型,对于每个分类模型根据信息增益比选择最好的样本特征进行***,直到所有训练样例都属于同一类,最后将生成的所有决策树分类模型组成随机森林,通过投票法输出驾驶风格类别K dri;所述驾驶风格类别K drive包括激进型、平和型、保守型三类:K drive=R f(C J) (2)。
- 根据权利要求2所述的自动驾驶汽车的认知***,其特征在于,所述层次化认知模块中,首先以自动驾驶汽车为中心节点,与中心节点存在耦合关系的节点和中心节点组成内层模块;其次,对内层模块的非中心节点进行重要度排序,依次寻找耦合系数最大的点组成中间层模块;然后,对中间层模块的节点进行重要度排序,依次寻找耦合系数最大的点组成外层模块;最后,由其它节点组成边缘层模块。
- 基于复杂网络的自动驾驶汽车认知***的认知方法,其特征在于,包括如下步骤:步骤1)提取纵向驾驶特征参数、横向驾驶特征参数和模式转移特征参数,构造驾驶风格特征矩阵C J,生成随机森林分类器R f,将驾驶风格特征矩阵C J输入随机森林分类 器R f,随机森林分类器R f的输出驾驶风格类别K drive,将驾驶风格识别为激进型、平和型、保守型三类;步骤2)构造时变复杂动态网络G作为复杂环境模型,用于描述复杂环境整体关联特征,进一步建立复杂环境模型中的节点动力学方程,再组合时变复杂动态网络G中所有节点的特征形成动态方程向量F(X)、时变复杂动态网络G中节点间的耦合矩阵P(t)和节点的内联向量H(X),建立时变复杂动态网络G的节点***动力学方程,用于描述复杂环境的动态特性;步骤3)构造复杂环境模型中节点的量g i、度k i、点权s i和重要度I(i)四个参数,并用正态分布图对节点进行差异化分析,实现节点差异化认知;步骤4)采用凝聚算法对复杂环境模型中节点进行层次划分,实现对自动驾驶汽车复杂环境的层次化、阶梯性认知;步骤5)依据熵理论的基本思想,用***熵和熵变对复杂环境模型的无序程度进行度量,描述整体风险及变化态势,实现对全局共性的状态认知。
- 根据权利要求9所述的认知方法,其特征在于,所述基于复杂网络的自动驾驶汽车认知***为权利要求2-8任一项所述的认知***。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US17/802,143 US20240190442A1 (en) | 2021-05-10 | 2022-01-07 | Complex network-based complex environment model, cognition system, and cognition method of autonomous vehicle |
DE112022000019.8T DE112022000019T5 (de) | 2021-05-10 | 2022-01-07 | Komplexes umgebungsmodell, kognitives system und kognitives verfahren eines selbstfahrenden fahrzeugs basierend auf einem komplexen netzwerk |
JP2022553145A JP7464236B2 (ja) | 2021-05-10 | 2022-01-07 | 複雑ネットワークによる自動運転車の複雑環境モデル、認知システム及び認知方法 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110504041.4 | 2021-05-10 | ||
CN202110504041.4A CN113406955B (zh) | 2021-05-10 | 2021-05-10 | 基于复杂网络的自动驾驶汽车复杂环境模型、认知***及认知方法 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022237212A1 true WO2022237212A1 (zh) | 2022-11-17 |
Family
ID=77678411
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2022/070671 WO2022237212A1 (zh) | 2021-05-10 | 2022-01-07 | 基于复杂网络的自动驾驶汽车复杂环境模型、认知***及认知方法 |
Country Status (5)
Country | Link |
---|---|
US (1) | US20240190442A1 (zh) |
JP (1) | JP7464236B2 (zh) |
CN (1) | CN113406955B (zh) |
DE (1) | DE112022000019T5 (zh) |
WO (1) | WO2022237212A1 (zh) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115622903A (zh) * | 2022-12-19 | 2023-01-17 | 中国人民解放军国防科技大学 | 一种基于网络结构的电信网节点重要度计算方法 |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113406955B (zh) * | 2021-05-10 | 2022-06-21 | 江苏大学 | 基于复杂网络的自动驾驶汽车复杂环境模型、认知***及认知方法 |
CN116811894B (zh) * | 2023-08-30 | 2023-11-21 | 北京理工大学 | 一种连续型驾驶风格识别方法、***及设备 |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101633359A (zh) * | 2008-07-24 | 2010-01-27 | 通用汽车环球科技运作公司 | 具有驾驶风格识别的自适应车辆控制*** |
CN103077603A (zh) * | 2012-06-06 | 2013-05-01 | 王晓原 | 基于动态人车环境协同推演的自由流状态汽车驾驶倾向性辨识*** |
CN106023344A (zh) * | 2016-06-06 | 2016-10-12 | 清华大学 | 基于驾驶模式转换概率的驾驶风格估计方法 |
US20190004526A1 (en) * | 2017-06-29 | 2019-01-03 | Continental Automotive Systems, Inc. | Propulsion efficient autonomous driving strategy |
CN109144076A (zh) * | 2018-10-31 | 2019-01-04 | 吉林大学 | 一种多车辆横纵向耦合协同控制***及控制方法 |
CN109927725A (zh) * | 2019-01-28 | 2019-06-25 | 吉林大学 | 一种具有驾驶风格学习能力的自适应巡航***及实现方法 |
CN110321954A (zh) * | 2019-07-03 | 2019-10-11 | 中汽研(天津)汽车工程研究院有限公司 | 适合国内人群的驾驶风格分类和识别方法及*** |
CN113406955A (zh) * | 2021-05-10 | 2021-09-17 | 江苏大学 | 基于复杂网络的自动驾驶汽车复杂环境模型、认知***及认知方法 |
Family Cites Families (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU2018347314A1 (en) * | 2017-10-10 | 2020-05-21 | The Government Of The United States Of America, As Represented By The Secretary Of The Navy | Method for identifying optimal vehicle paths when energy is a key metric or constraint |
US11378956B2 (en) * | 2018-04-03 | 2022-07-05 | Baidu Usa Llc | Perception and planning collaboration framework for autonomous driving |
CN108725453A (zh) * | 2018-06-11 | 2018-11-02 | 南京航空航天大学 | 基于驾驶员模型和操纵逆动力学的人机共驾控制***及其切换模式 |
CN109829577B (zh) * | 2019-01-17 | 2021-10-01 | 北京交通大学 | 基于深度神经网络结构模型的轨道列车运行状态预测方法 |
CN109948781A (zh) * | 2019-03-21 | 2019-06-28 | 中国人民解放军国防科技大学 | 用于自动驾驶车辆的连续动作在线学习控制方法及*** |
CN110160804B (zh) * | 2019-05-31 | 2020-07-31 | 中国科学院深圳先进技术研究院 | 一种自动驾驶车辆的测试方法、装置及*** |
CN111539112B (zh) * | 2020-04-27 | 2022-08-05 | 吉林大学 | 一种用于自动驾驶车快速查找交通对象的场景建模方法 |
CN111897217B (zh) * | 2020-07-20 | 2022-03-11 | 清华大学 | 一种模型预测控制器的时域分解加速方法 |
CN112015842B (zh) * | 2020-09-02 | 2024-02-27 | 中国科学技术大学 | 自行车轨迹预测的自动驾驶车辆风险评估方法及*** |
CN112437501B (zh) * | 2020-10-19 | 2022-11-18 | 江苏大学 | 一种基于交通语义和博弈论的多传感器超视距自组网方法 |
-
2021
- 2021-05-10 CN CN202110504041.4A patent/CN113406955B/zh active Active
-
2022
- 2022-01-07 WO PCT/CN2022/070671 patent/WO2022237212A1/zh active Application Filing
- 2022-01-07 DE DE112022000019.8T patent/DE112022000019T5/de active Pending
- 2022-01-07 US US17/802,143 patent/US20240190442A1/en active Pending
- 2022-01-07 JP JP2022553145A patent/JP7464236B2/ja active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101633359A (zh) * | 2008-07-24 | 2010-01-27 | 通用汽车环球科技运作公司 | 具有驾驶风格识别的自适应车辆控制*** |
US20100023223A1 (en) * | 2008-07-24 | 2010-01-28 | Gm Global Technology Operations, Inc. | Adaptive vehicle control system with driving style recognition |
CN103077603A (zh) * | 2012-06-06 | 2013-05-01 | 王晓原 | 基于动态人车环境协同推演的自由流状态汽车驾驶倾向性辨识*** |
CN106023344A (zh) * | 2016-06-06 | 2016-10-12 | 清华大学 | 基于驾驶模式转换概率的驾驶风格估计方法 |
US20190004526A1 (en) * | 2017-06-29 | 2019-01-03 | Continental Automotive Systems, Inc. | Propulsion efficient autonomous driving strategy |
CN109144076A (zh) * | 2018-10-31 | 2019-01-04 | 吉林大学 | 一种多车辆横纵向耦合协同控制***及控制方法 |
CN109927725A (zh) * | 2019-01-28 | 2019-06-25 | 吉林大学 | 一种具有驾驶风格学习能力的自适应巡航***及实现方法 |
CN110321954A (zh) * | 2019-07-03 | 2019-10-11 | 中汽研(天津)汽车工程研究院有限公司 | 适合国内人群的驾驶风格分类和识别方法及*** |
CN113406955A (zh) * | 2021-05-10 | 2021-09-17 | 江苏大学 | 基于复杂网络的自动驾驶汽车复杂环境模型、认知***及认知方法 |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115622903A (zh) * | 2022-12-19 | 2023-01-17 | 中国人民解放军国防科技大学 | 一种基于网络结构的电信网节点重要度计算方法 |
CN115622903B (zh) * | 2022-12-19 | 2023-04-07 | 中国人民解放军国防科技大学 | 一种基于网络结构的电信网节点重要度计算方法 |
Also Published As
Publication number | Publication date |
---|---|
DE112022000019T5 (de) | 2023-01-26 |
CN113406955B (zh) | 2022-06-21 |
JP2023528114A (ja) | 2023-07-04 |
JP7464236B2 (ja) | 2024-04-09 |
US20240190442A1 (en) | 2024-06-13 |
CN113406955A (zh) | 2021-09-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2022237212A1 (zh) | 基于复杂网络的自动驾驶汽车复杂环境模型、认知***及认知方法 | |
Kashinath et al. | Review of data fusion methods for real-time and multi-sensor traffic flow analysis | |
Sheng et al. | Graph-based spatial-temporal convolutional network for vehicle trajectory prediction in autonomous driving | |
US11899411B2 (en) | Hybrid reinforcement learning for autonomous driving | |
EP4018376A1 (en) | Performance testing for robotic systems | |
JP2022016419A (ja) | 軌跡予測方法及び装置 | |
Malik et al. | Driving pattern profiling and classification using deep learning | |
Masmoudi et al. | A reinforcement learning framework for video frame-based autonomous car-following | |
CN112613225B (zh) | 一种基于神经网络元胞传输模型的交叉***通状态预测方法 | |
CN102054176A (zh) | 用计算机对运动目标的场景图像建立语义场景模型的方法 | |
CN114169682A (zh) | 一种驾驶能力过程性评价方法和*** | |
CN114566052B (zh) | 一种基于车流方向判别高速公路车流监控设备转动的方法 | |
Chu et al. | A review of driving style recognition methods from short-term and long-term perspectives | |
Liu et al. | An inverse reinforcement learning approach for customizing automated lane change systems | |
Vlahogianni et al. | Model free identification of traffic conditions using unmanned aerial vehicles and deep learning | |
Li et al. | Driver-specific risk recognition in interactive driving scenarios using graph representation | |
CN117217314A (zh) | 一种基于元数据驱动与因果分析理论的驾驶情境推理方法 | |
US20240001942A1 (en) | Performance testing for robotic systems | |
Xue et al. | Traffic status evolution trend prediction based on congestion propagation effects under rainy weather | |
Yu et al. | Longitudinal wind field prediction based on DDPG | |
Khan et al. | Deep ensemble for rotorcraft attitude prediction | |
Jose et al. | A New approach for accurate path Prediction using multiple Prediction system | |
Zhang et al. | Road condition recognition in self-driving cars based on classification and regression tree | |
Islam et al. | Elevating Driver Behavior Understanding with RKnD: A Novel Probabilistic Feature Engineering Approach | |
Sarath | Deep NN-Based Autonomous Vehicle Framework for Route Selection and Traffic Control Using Golden Levy Optimization |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 17802143 Country of ref document: US |
|
ENP | Entry into the national phase |
Ref document number: 2022553145 Country of ref document: JP Kind code of ref document: A |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 22806190 Country of ref document: EP Kind code of ref document: A1 |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 22806190 Country of ref document: EP Kind code of ref document: A1 |