CN113516846A - Vehicle lane change behavior prediction model construction, prediction and early warning method and system - Google Patents

Vehicle lane change behavior prediction model construction, prediction and early warning method and system Download PDF

Info

Publication number
CN113516846A
CN113516846A CN202110702854.4A CN202110702854A CN113516846A CN 113516846 A CN113516846 A CN 113516846A CN 202110702854 A CN202110702854 A CN 202110702854A CN 113516846 A CN113516846 A CN 113516846A
Authority
CN
China
Prior art keywords
vehicle
target vehicle
vehicles
lane change
data
Prior art date
Legal status (The legal status 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 status listed.)
Granted
Application number
CN202110702854.4A
Other languages
Chinese (zh)
Other versions
CN113516846B (en
Inventor
惠飞
贾硕
魏诚
房山
靳少杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Changan University
Original Assignee
Changan University
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 Changan University filed Critical Changan University
Priority to CN202110702854.4A priority Critical patent/CN113516846B/en
Publication of CN113516846A publication Critical patent/CN113516846A/en
Application granted granted Critical
Publication of CN113516846B publication Critical patent/CN113516846B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems
    • G08G1/161—Decentralised systems, e.g. inter-vehicle communication
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems
    • G08G1/167—Driving aids for lane monitoring, lane changing, e.g. blind spot detection
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/02—Services making use of location information
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/30—Services specially adapted for particular environments, situations or purposes
    • H04W4/40—Services specially adapted for particular environments, situations or purposes for vehicles, e.g. vehicle-to-pedestrians [V2P]
    • H04W4/46—Services specially adapted for particular environments, situations or purposes for vehicles, e.g. vehicle-to-pedestrians [V2P] for vehicle-to-vehicle communication [V2V]

Abstract

The invention discloses a method and a system for constructing, predicting and early warning a vehicle lane change behavior prediction model, which utilize a vehicle-mounted terminal to acquire vehicle running state data in real time and carry out data communication among vehicles; by combining the game theory and the deep learning technology, the dynamic interaction of different vehicles in the driving process is obtained from the data, the information of the driving environment around the vehicles is analyzed, whether the current environment is suitable for lane changing of the vehicles is judged by using the game theory, the lane changing intention of a driver is further quantified, the running data of the vehicles is identified and predicted by using the deep learning algorithm, and when the running data begin to meet the characteristics of lane changing, the method can obtain the prediction result that the vehicles are executing lane changing operation in a short time, and can also send the predicted lane changing behavior to vehicle-mounted terminals of the surrounding vehicles in a workshop communication mode to play a role in early warning.

Description

Vehicle lane change behavior prediction model construction, prediction and early warning method and system
Technical Field
The invention belongs to the technical field of auxiliary driving, and relates to a vehicle lane change behavior prediction model construction, prediction and early warning method and system, in particular to a method for predicting vehicle lane change behavior by combining a game theory and a deep learning algorithm, and modules for data acquisition, workshop communication, warning display and the like required by an early warning system and required by the prediction method.
Background
The driving behavior of drivers is closely related to road traffic safety, and most of traffic accidents are caused by improper driving operation. Of all the operations by the driver, the lateral operations (right and left lane changes and turning) of the vehicle have the greatest influence on the stability of the traffic flow, and cause the most accidents. The existing Advanced Driver Assistance System (ADAS) integrates a Blind-Spot-Warning System (Blind-Spot-Warning System) and a Lane Departure Warning System (Lane Departure Warning System), which can reduce the probability of accidents caused by transverse operation to a certain extent, however, the normal operation of these Systems depends on the Driver to correctly use a turn signal, and in the actual life, some drivers do not use the turn signal to warn surrounding vehicles and passersby before changing lanes or turning lanes, so the functions of these Systems are limited, and the problem of frequent accidents caused by transverse operation cannot be solved. Therefore, the mathematical model is used for simulating the perception of the driver to the surrounding driving environment and the expected driving state in the driving process, and the vehicle running data collected by the vehicle-mounted terminal is used for accurately and efficiently predicting the lateral operation such as lane change to be carried out by the driver, so that the purpose of warning the vehicles which are possibly affected around before the vehicle motion state is changed can be achieved, the occurrence of accidents is reduced, and the road traffic safety is guaranteed.
Although recent research starts to identify and predict driving behaviors, on one hand, data sets used in some research are historical data collected in advance, so that the constructed algorithm framework can only be used for identifying driving operations which have already occurred, and can not be used for predicting the driving behaviors of a driver in the future in real time; on one hand, a mathematical model (a hidden Markov model, a game theory model and the like) is used for predicting real-time lane change intention at present, although the potential possibility can be predicted before the lane change action occurs, the prediction mode can only draw the conclusion that the current driving environment is suitable for the lane change of the vehicle, and whether the driver performs according to the intention is unknown, so that the prediction accuracy is difficult to guarantee; on the other hand, the deep learning algorithm has unique advantages in predicting continuous time sequence data similar to driving behavior data, and can achieve an accurate prediction effect through a large number of historical data segment training algorithm models, but the algorithm can often obtain accurate prediction accuracy within a period of time after the lane changing behavior starts, the real-time performance of behavior prediction is difficult to guarantee, and in practical traffic application, the lane changing behavior of a vehicle is predicted and surrounding vehicles and people are warned every 0.1 second in advance, so that accidents can be reduced to a great extent.
Disclosure of Invention
Aiming at the immaturity of the prior art, the invention provides a vehicle lane change behavior prediction model construction, prediction and early warning method and system, which consider the viewpoint of real-time performance and accuracy of a driver in the lane change behavior prediction process, reduce the time sequence data length required by an algorithm to obtain a correct prediction result on the premise of ensuring higher prediction accuracy, namely lower misjudgment rate, namely obtain the correct prediction result in a shorter time, and send early warning information to surrounding vehicles in time, and aim to solve the problem of prediction of the lane change behavior of the driver.
In order to achieve the purpose, the invention adopts the following technical scheme:
a method for constructing a prediction model of vehicle lane change behavior, which is used for predicting the vehicle lane change behavior, comprises the following steps:
step 1, collecting vehicle running state data:
a vehicle-mounted terminal is arranged in each vehicle, and the vehicle-mounted terminal integrates a CAN bus module to collect vehicle running data, a GPS module to acquire positioning data of the vehicles, and a communication module to realize data communication between the vehicles and a display module to display the collected vehicle running state data; the vehicle travel data includes a speed, an acceleration, and a steering angle of the vehicle;
step 2, data communication between vehicles:
each vehicle sends the vehicle running state data to other vehicles through the communication module of the vehicle, receives the vehicle running state data of other vehicles, and randomly selects one vehicle as a target vehicle;
step 3, obtaining the lane change intention probability of the lane change behavior of the target vehicle:
after receiving vehicle running state data of other vehicles, the target vehicle establishes a game matrix by combining the vehicle running state data of the target vehicle and utilizing a game theory, and then solves the game matrix to obtain the lane change intention probability of the target vehicle;
and 4, establishing a deep learning algorithm model, taking the vehicle driving data and the lane change intention probability as driving sample data in the lane change process of the vehicle, inputting a training sample set consisting of the driving sample data into the deep learning algorithm model, taking whether the vehicle is subjected to lane change as the output of the deep learning algorithm model, training the deep learning algorithm model, and taking the trained deep learning algorithm model as a vehicle lane change behavior prediction model.
The invention also comprises the following technical characteristics:
specifically, in the step 3, a game matrix is established by using a game theory, and when a target vehicle is influenced by a front vehicle with a slow speed, the strategy of the target vehicle comprises lane changing or lane changing not; meanwhile, the target vehicle can influence the running of the rear vehicle close to the lane, so that the rear vehicle can give way or not give way by two strategies; the gaming matrix of the target vehicle and the rear workshop therefore includes the following:
the lane change is required for the target vehicle, and under the condition that the rear vehicle gives way to the target vehicle: target vehicle revenue is USVCThe benefit of the rear vehicle is UV1A；
The lane change is required for the target vehicle, and the rear vehicle does not give way to the target vehicle: target vehicle revenue is USVCThe benefit of the rear vehicle is UV1R；
And when the target vehicle does not require lane changing and the rear vehicle gives way to the target vehicle: target vehicle revenue is USVSThe benefit of the rear vehicle is UV1A；
The target vehicle does not require lane changing, and the rear vehicle does not give way to the target vehicle: target vehicle revenue is USVSThe benefit of the rear vehicle is UV1R。
Specifically, each revenue function under different strategies is a safety revenue UsafetyAnd space profit UspaceThe composition is as follows:
U＝αUsafety+βUspacewherein β is a Sigmoid function, α ═ 1- β;
wherein:
Figure BDA0003130840890000031
Figure BDA0003130840890000032
in the above formula, T (T) is the headway at time T, Tmin(t) minimum safe headway, D (t) distance to front, Dmin(t) is the minimum safe distance.
Specifically, the minimum safe distance between the target vehicle and the rear vehicle is as follows:
Figure BDA0003130840890000033
wherein, aV1For rear vehicle acceleration, asvIs the target vehicle acceleration, vv1Is the rear vehicle speed, vsvIs the target vehicle speed;
minimum safe headway of target vehicle and back car:
Tmin(t)＝min(Tinitial,Ta)
wherein, TinitialThe time interval between the rear vehicle and the front vehicle at the time T, TaIs a fixed value and is set to 3.
Specifically, in the step 3, solving the game matrix to obtain the lane change intention probability of the target vehicle includes:
firstly, the target vehicle judges the strategy of the maximum profit of the target vehicle according to the game matrix, namely lane changing or lane not changing, then the rear vehicle selects the reaction of the maximum profit of the target vehicle under the current condition according to the decision of the target vehicle, namely lane giving or lane not giving, and then the reaction is fed back to the target vehicle, and the circulation is continued until both game parties are satisfied, and the final balanced state is reached;
according to the game matrix, under the equilibrium state, the decision gain of the target vehicle is UsvAnd the other decision benefit is instead Usv*And then the quantifiable game result, namely the lane change probability is as follows:
Figure BDA0003130840890000041
wherein Pro (t) is the game result and is the lane change probability of the target vehicle.
Specifically, the deep learning algorithm model consists of a long-short term memory network LSTM and a convolutional neural network CNN; in the deep learning algorithm model, the number of hidden layer nodes is set to be 100, the learning rate is 0.001, the step length is 8, and in the CNN, the convolution kernel is a 3 × 3 matrix and the convolution step length is 1.
A prediction and early warning method for a lane changing behavior of a vehicle comprises the following steps:
step one, collecting vehicle driving state data:
a vehicle-mounted terminal is arranged in each vehicle, and the vehicle-mounted terminal integrates a CAN bus module to collect vehicle running data, a GPS module to acquire positioning data of the vehicles, and a communication module to realize data communication between the vehicles and a display module to display the collected vehicle running state data; the vehicle travel data includes a speed, an acceleration, and a steering angle of the vehicle;
step two, data communication between vehicles:
each vehicle transmits its vehicle driving state data to other vehicles through its communication module and receives the vehicle driving state data of the other vehicles;
step three, obtaining the lane change intention probability of the lane change behavior of the target vehicle by the method in the step 3 in the vehicle lane change behavior prediction model construction method;
step four, taking the driving state data of the target vehicle and the lane change intention probability as input, importing the data into the trained deep learning algorithm model, and predicting whether the target vehicle changes lanes or not; and if the target vehicle is predicted to change the lane, displaying early warning information on vehicle-mounted terminals of other vehicles.
The utility model provides a vehicle lane change behavior prediction early warning system which characterized in that includes:
the vehicle running state data acquisition module comprises a vehicle-mounted terminal arranged in each vehicle, wherein the vehicle-mounted terminal integrates a CAN bus module to acquire vehicle running data, a GPS module to acquire positioning data of the vehicles, and a communication module to realize data communication between the vehicles and a display module to display the acquired vehicle running state data;
the inter-vehicle data communication module is used for transmitting the vehicle driving state data of each vehicle to other vehicles and receiving the vehicle driving state data of other vehicles;
the lane change intention probability acquisition module is used for establishing a game matrix by combining the driving state data of the vehicle after the target vehicle receives the driving state data of other vehicles and utilizing a game theory after the target vehicle receives the driving state data of the other vehicles, and then solving the game matrix to acquire the lane change intention probability of the target vehicle;
the target vehicle lane change prediction and early warning module is used for importing the driving state data of the target vehicle and the lane change intention probability into a trained deep learning algorithm model to predict whether the target vehicle changes lanes or not; and if the target vehicle is predicted to change the lane, displaying early warning information on vehicle-mounted terminals of other vehicles.
Compared with the prior art, the invention has the beneficial technical effects that:
1. the invention can utilize the vehicle-mounted terminal to acquire, send and receive the driving state of the vehicle in real time. 2. The invention can use the vehicle running state data to estimate the intention of the vehicle to change the lane. 3. The invention can combine the lane changing intention with the driving state data to accurately predict the lane changing behavior of the vehicle in time. 4. The invention can send the predicted lane change behavior to the vehicle-mounted terminals of the surrounding vehicles in a workshop communication mode, thereby playing a role of early warning.
Drawings
FIG. 1 is a schematic diagram of a channel change prediction and early warning system implementation process of the present invention;
FIG. 2 is a schematic diagram of a vehicle terminal design framework;
FIG. 3 is a schematic diagram of a vehicle terminal hardware device;
FIG. 4 is a schematic illustration of a plant communication system operational interface;
FIG. 5 is a schematic view of a lane-change behavior scenario of a vehicle;
FIG. 6 is a diagram illustrating the prediction accuracy of a prediction model;
fig. 7 is a schematic diagram of warning information display.
Detailed Description
In order to solve the problems of low prediction accuracy and long prediction time in the prior art, dynamic interaction of different vehicles in the driving process is obtained from data by combining a game theory and a deep learning technology, the information of the driving environment around the vehicles is analyzed, whether the current environment is suitable for lane changing of the vehicles is judged by using a game theory, the lane changing intention of a driver is further quantized, the running data of the vehicles are identified and predicted by using a deep learning algorithm under the condition that the lane changing intention of the driver is determined, and when the running data start to meet the characteristics of lane changing, the whole model can obtain the prediction result of the lane changing operation of the vehicles in a short time; the invention can also send the prediction result of the vehicle about to change the lane to the surrounding vehicles by using the vehicle-mounted terminal, thereby achieving the effect of early warning, avoiding accidents to the greatest extent and solving the problem that the prior art can not send early warning to the surrounding vehicles in time.
The following definitions or conceptual connotations relating to the present invention are provided for illustration:
game theory: the Game Theory, also called the Game Theory, the Game office Theory, etc., is a new branch of modern mathematics and an important discipline of operational research. Generally consists of several elements including: participants, actions, information, policies, benefits, outcomes, balances, etc.
Game matrix: the set of policies of a plurality of participants participating in a game can be represented by a matrix or block diagram, which is called a game matrix.
The revenue function: each participant in the game calculates a function of the benefit available based on the type of participant and the selected action when the participant in the game plays.
A safety revenue function: the calculation function of the profit under driving safety is available to each participant participating in the game.
A spatial gain function: the revenue calculation function in terms of travel space available to each participant participating in the game.
The following describes in detail specific embodiments of the present invention. It should be understood that the detailed description and specific examples, while indicating the present invention, are given by way of illustration and explanation only, not limitation.
The invention provides a method for constructing a prediction model of a lane change behavior of a vehicle, which comprises the following steps of:
step one, collecting vehicle driving state data:
a vehicle-mounted terminal is arranged in each vehicle, and the vehicle-mounted terminal integrates a CAN bus module to collect vehicle running data, a GPS module to acquire vehicle positioning data, and a communication module to realize inter-vehicle data communication and a display module to display the collected vehicle running state data, as shown in FIGS. 2 to 4, FIG. 2 is a schematic diagram of a vehicle-mounted terminal design framework, FIG. 3 is a schematic diagram of vehicle-mounted terminal hardware equipment, and FIG. 4 is a schematic diagram of a vehicle-to-vehicle communication system display module operation interface; the vehicle travel data includes a speed, an acceleration, and a steering angle of the vehicle;
further, in the first step, the vehicle-mounted terminal adopts a design scheme of an MCU main control board and a communication module, the main control board adopts an NXP IMX6 development board, a CAN processing chip and an external interface are integrated, and the GPS module adopts an NEO7N module; CAN collect CAN bus data and GPS data in the running process of the vehicle.
Step two, data communication between vehicles:
each vehicle sends the vehicle running state data thereof to other vehicles through the communication module of the vehicle, receives the vehicle running state data of other vehicles, and arbitrarily selects one vehicle as a target vehicle, wherein a vehicle in fig. 5 is the target vehicle;
further, in the second step, the inter-vehicle communication equipment is a Zhongxing ZM 8350C-V2X module integrated on the MCU main control board, and completes the tasks of sending and receiving data.
Step three, obtaining the lane change intention probability of the lane change behavior of the target vehicle:
after receiving vehicle running state data of other vehicles, the target vehicle establishes a game matrix by combining the vehicle running state data of the target vehicle and utilizing a game theory, and then solves the game matrix to obtain the lane change intention probability of the target vehicle; further, in the third step, the game theory considers the game among drivers as a non-cooperative static game under complete information.
More specifically, in the third step, a game theory is utilized to establish a game matrix, and when the target vehicle is influenced by a front vehicle with a slow speed, the strategy of the target vehicle comprises lane changing or lane changing not; meanwhile, the target vehicle can influence the running of the rear vehicle close to the lane, so that the rear vehicle can give way or not give way by two strategies; the game matrix for the target vehicle and the rear shop is thus as follows:
TABLE 1 Game matrix for target vehicles and post-shop
Figure BDA0003130840890000071
The gaming matrix includes the following cases:
the lane change is required for the target vehicle, and under the condition that the rear vehicle gives way to the target vehicle: target vehicle revenue is USVCThe benefit of the rear vehicle is UV1A；
For target vehicleThe lane change is required, and the rear vehicle does not give way to the target vehicle: target vehicle revenue is USVCThe benefit of the rear vehicle is UV1R；
And when the target vehicle does not require lane changing and the rear vehicle gives way to the target vehicle: target vehicle revenue is USVSThe benefit of the rear vehicle is UV1A；
The target vehicle does not require lane changing, and the rear vehicle does not give way to the target vehicle: target vehicle revenue is USVSThe benefit of the rear vehicle is UV1R。
Each income function under different strategies is composed of a safety income UsafetyAnd space profit UspaceThe composition is as follows:
U＝αUsafety+βUspacewherein β is a Sigmoid function, α ═ 1- β;
wherein:
Figure BDA0003130840890000072
Figure BDA0003130840890000073
in the above formula, T (T) is the headway at time T, Tmin(t) minimum safe headway, D (t) distance to front, Dmin(t) is the minimum safe distance.
Minimum safe distance of target vehicle to rear vehicle:
Figure BDA0003130840890000081
wherein, aV1For rear vehicle acceleration, asvIs the target vehicle acceleration, vv1Is the rear vehicle speed, vsvIs the target vehicle speed;
minimum safe headway of target vehicle and back car:
Tmin(t)＝min(Tinitial,Ta)
wherein, TinitialThe time interval between the rear vehicle and the front vehicle at the time T, TaIs a fixed value and is set to 3.
In the third step, solving the game matrix to obtain the lane change intention probability of the target vehicle comprises the following steps:
firstly, the target vehicle judges the strategy of the maximum profit of the target vehicle according to the game matrix, namely lane changing or lane not changing, then the rear vehicle selects the reaction of the maximum profit of the target vehicle under the current condition according to the decision of the target vehicle, namely lane giving or lane not giving, and then the reaction is fed back to the target vehicle, and the circulation is continued until both game parties are satisfied, and the final balanced state is reached;
according to the game matrix, under the equilibrium state, the decision gain of the target vehicle is UsvAnd the other decision benefit is instead Usv*And then the quantifiable game result, namely the lane change probability is as follows:
Figure BDA0003130840890000082
wherein Pro (t) is a game result and is also the lane change probability of the target vehicle; if the game is passed, the decision of the target vehicle is lane change, and the lane change probability of the target vehicle can be calculated as follows:
Figure BDA0003130840890000083
and step four, establishing a deep learning algorithm model, taking the vehicle driving data and the lane change intention probability as driving sample data in the lane change process of the vehicle, inputting a training sample set consisting of the driving sample data into the deep learning algorithm model, taking whether the vehicle is subjected to lane change as the output of the deep learning algorithm model, training the deep learning algorithm model, and taking the trained deep learning algorithm model as a vehicle lane change behavior prediction model. The deep learning algorithm model consists of a long-short term memory network LSTM and a convolutional neural network CNN; wherein the network structure of the LSTM can be described by three control gates and a memory state, which can be represented by the following formula:
an input gate:
Figure BDA0003130840890000091
Figure BDA0003130840890000092
forget the door:
Figure BDA0003130840890000093
Figure BDA0003130840890000094
an output gate:
Figure BDA0003130840890000095
Figure BDA0003130840890000096
a storage unit:
Figure BDA0003130840890000097
Figure BDA0003130840890000098
and (3) outputting:
Figure BDA0003130840890000099
in the above-mentioned equation, the equation,
Figure BDA00031308408900000910
the input data for each moment consists of the running state of the vehicle (horizontal/longitudinal speed, horizontal/longitudinal acceleration and steering angle) and the result of the game theory model.
Figure BDA00031308408900000911
Indicating the memory of the information at the previous moment.
Figure BDA00031308408900000912
Is composed of input data and cell states, each representing a computation of a corresponding part of the network, including an input gate, a forgetting gate, an output gate, and a memory cell (cell).
Figure BDA0003130840890000101
Respectively, the outputs of the various parts after different activation functions. Omegail、ωcl、
Figure BDA0003130840890000102
ωiw、ωcwRespectively, the weight matrix of the network. During the learning process of the network, their values gradually tend to the optimal weight. The result of this module is
Figure BDA0003130840890000103
It represents the output of blocks.
Importing the data processed by the LSTM into a CNN assembly, and further extracting features from the result; the mathematical model of CNN is represented by the following function:
and (3) rolling layers: the function of the convolution layer is to perform feature extraction on input data, and each element forming a convolution kernel corresponds to a weight coefficient and a deviation value.
Figure BDA0003130840890000104
A pooling layer: after feature extraction is performed on the convolutional layer, the output features are transmitted to the pooling layer for feature selection and information filtering.
c＝[c1，c2，…，cn-k+1]
Figure BDA0003130840890000105
Wherein c isiFor the convolution result, W ∈ R and b ∈ R are parameters in the convolution process and represent convolution kernels and biases.
c is equal to R and
Figure BDA0003130840890000106
is a simple expression of the pooling process; and after convolution and pooling operations are carried out, outputting a prediction result of whether the vehicle changes the lane or not by using the whole model by using the full communication layer.
More specifically, in the present embodiment, in the deep learning algorithm model, the number of hidden layer nodes is set to 100, the learning rate is 0.001, the step size is 8, and in the CNN, the convolution kernel is a 3 × 3 matrix, and the convolution step size is 1.
In this embodiment, the sample data is a driving data segment which is acquired in advance, segmented and labeled manually in the lane change process of the vehicle, the data dimensions include lateral/longitudinal speed, lateral/longitudinal acceleration, steering angle and game results, the length of each sample segment is 20, the data frequency is 10Hz, 1659 lane change samples and 986 lane change-free samples are used together to train the network, the number of the right lane change samples is 509, and the number of the left lane change samples is 1150.
In this embodiment, the trained network prediction result is shown in fig. 6, and the accuracy can reach about 94%.
The invention also provides a vehicle lane change behavior prediction and early warning method, which comprises the following steps:
step one, collecting vehicle driving state data:
a vehicle-mounted terminal is arranged in each vehicle, and the vehicle-mounted terminal integrates a CAN bus module to collect vehicle running data, a GPS module to acquire positioning data of the vehicles, and a communication module to realize data communication between the vehicles and a display module to display the collected vehicle running state data; the vehicle travel data includes a speed, an acceleration, and a steering angle of the vehicle;
step two, data communication between vehicles:
each vehicle transmits its vehicle driving state data to other vehicles through its communication module and receives the vehicle driving state data of the other vehicles;
step three, obtaining the lane change intention probability of the lane change behavior of the target vehicle by the method in the step 3 in the vehicle lane change behavior prediction model construction method;
step four, taking the driving state data of the target vehicle and the lane change intention probability as input, importing the data into the trained deep learning algorithm model, and predicting whether the target vehicle changes lanes or not; if it is predicted that the target vehicle will make a lane change, the warning information is displayed on the in-vehicle terminals of the other vehicles, as shown in fig. 7.
The invention also provides a vehicle lane change behavior prediction and early warning system, which comprises:
the vehicle running state data acquisition module comprises a vehicle-mounted terminal arranged in each vehicle, wherein the vehicle-mounted terminal integrates a CAN bus module to acquire vehicle running data, a GPS module to acquire positioning data of the vehicles, and a communication module to realize data communication between the vehicles and a display module to display the acquired vehicle running state data;
the inter-vehicle data communication module is used for transmitting the vehicle driving state data of each vehicle to other vehicles and receiving the vehicle driving state data of other vehicles;
the lane change intention probability acquisition module is used for establishing a game matrix by combining the driving state data of the vehicle after the target vehicle receives the driving state data of other vehicles and utilizing a game theory after the target vehicle receives the driving state data of the other vehicles, and then solving the game matrix to acquire the lane change intention probability of the target vehicle;
the target vehicle lane change prediction and early warning module is used for importing the driving state data of the target vehicle and the lane change intention probability into a trained deep learning algorithm model to predict whether the target vehicle changes lanes or not; and if the target vehicle is predicted to change the lane, displaying early warning information on vehicle-mounted terminals of other vehicles.
And (3) experimental verification:
compared with the prior art and experimental effects, in the aspects of the prediction accuracy and prediction time of the algorithm, the prior art and the technique provided by the invention are tested by using real data, and the test results are shown in the following table 1:
TABLE 1 test results
Figure BDA0003130840890000121
The test results show that the accuracy of the lane change behavior prediction algorithm based on the game theory is improved, and the prediction time based on the deep learning algorithm is reduced.
The invention provides a complete behavior prediction process, which comprises data acquisition, workshop communication, a prediction algorithm and an early warning module, and realizes accurate prediction and early warning of lane change behavior of a driver theoretically and practically. The data acquisition module ensures the data source under the real condition of the invention, and achieves the purpose of acquiring the vehicle operation data in real time; the vehicle-to-vehicle communication module ensures the effectiveness of the vehicle networking and the feasibility of the invention, and can complete the data transmission between vehicles and the sending of early warning information; the prediction algorithm combines the game theory and deep learning, and reduces the prediction time as much as possible on the premise of ensuring the prediction precision; the early warning module utilizes workshop communication to send the prediction result to display device on, can reach the purpose of reminding other vehicle drivers.

Claims (8)

1. A method for constructing a prediction model of a lane change behavior of a vehicle, wherein the model is used for predicting the lane change behavior of the vehicle, and is characterized by comprising the following steps:
step 1, collecting vehicle running state data:
a vehicle-mounted terminal is arranged in each vehicle, and the vehicle-mounted terminal integrates a CAN bus module to collect vehicle running data, a GPS module to acquire positioning data of the vehicles, and a communication module to realize data communication between the vehicles and a display module to display the collected vehicle running state data; the vehicle travel data includes a speed, an acceleration, and a steering angle of the vehicle;
step 2, data communication between vehicles:
each vehicle sends the vehicle running state data to other vehicles through the communication module of the vehicle, receives the vehicle running state data of other vehicles, and randomly selects one vehicle as a target vehicle;
step 3, obtaining the lane change intention probability of the lane change behavior of the target vehicle:
after receiving vehicle running state data of other vehicles, the target vehicle establishes a game matrix by combining the vehicle running state data of the target vehicle and utilizing a game theory, and then solves the game matrix to obtain the lane change intention probability of the target vehicle;
and 4, establishing a deep learning algorithm model, taking the vehicle driving data and the lane change intention probability as driving sample data in the lane change process of the vehicle, inputting a training sample set consisting of the driving sample data into the deep learning algorithm model, taking whether the vehicle is subjected to lane change as the output of the deep learning algorithm model, training the deep learning algorithm model, and taking the trained deep learning algorithm model as a vehicle lane change behavior prediction model.
2. The method for constructing the vehicle lane change behavior prediction model according to claim 1, wherein in the step 3, a game matrix is established by using a game theory, and when a target vehicle is influenced by a front vehicle with a slow speed, the strategy of the target vehicle comprises lane change or lane change not; meanwhile, the target vehicle can influence the running of the rear vehicle close to the lane, so that the rear vehicle can give way or not give way by two strategies; the gaming matrix of the target vehicle and the rear workshop therefore includes the following:
the lane change is required for the target vehicle, and under the condition that the rear vehicle gives way to the target vehicle: target vehicle revenue is USVCThe benefit of the rear vehicle is UV1A；
The target vehicle is required to be changed, and the rear vehicle does not give the target vehicleUnder the condition of lane yielding: target vehicle revenue is USVCThe benefit of the rear vehicle is UV1R；
And when the target vehicle does not require lane changing and the rear vehicle gives way to the target vehicle: target vehicle revenue is USVSThe benefit of the rear vehicle is UV1A；
The target vehicle does not require lane changing, and the rear vehicle does not give way to the target vehicle: target vehicle revenue is USVSThe benefit of the rear vehicle is UV1R。
3. The method according to claim 2, wherein each revenue function under different strategies is a safety revenue UsafetyAnd space profit UspaceThe composition is as follows:
U＝αUsafety+βUspacewherein β is a Sigmoid function, α ═ 1- β;
wherein:
Figure FDA0003130840880000021
in the above formula, T (T) is the headway at time T, Tmin(t) minimum safe headway, D (t) distance to front, Dmin(t) is the minimum safe distance.
4. The vehicle lane change behavior prediction model construction method according to claim 3, wherein the minimum safe distance between the target vehicle and the following vehicle is:
Figure FDA0003130840880000022
wherein, aV1For rear vehicle acceleration, asvIs the target vehicle acceleration, vv1Is the rear vehicle speed, vsvIs the target vehicle speed;
minimum safe headway of target vehicle and back car:
Tmin(t)＝min(Tinitial,Ta)
wherein, TinitialThe time interval between the rear vehicle and the front vehicle at the time T, TaIs a fixed value and is set to 3.
5. The method for constructing the prediction model of the lane changing behavior of the vehicle according to claim 4, wherein in the step 3, solving the game matrix to obtain the probability of the lane changing intention of the target vehicle comprises:
firstly, the target vehicle judges the strategy of the maximum profit of the target vehicle according to the game matrix, namely lane changing or lane not changing, then the rear vehicle selects the reaction of the maximum profit of the target vehicle under the current condition according to the decision of the target vehicle, namely lane giving or lane not giving, and then the reaction is fed back to the target vehicle, and the circulation is continued until both game parties are satisfied, and the final balanced state is reached;
according to the game matrix, under the equilibrium state, the decision gain of the target vehicle is UsvAnd the other decision benefit is instead Usv*And then the quantifiable game result, namely the lane change probability is as follows:
Figure FDA0003130840880000031
wherein Pro (t) is the game result and is the lane change probability of the target vehicle.
6. The vehicle lane change behavior prediction model construction method according to claim 1, wherein the deep learning algorithm model is composed of a long-short term memory network LSTM and a convolutional neural network CNN; in the deep learning algorithm model, the number of hidden layer nodes is set to be 100, the learning rate is 0.001, the step length is 8, and in the CNN, the convolution kernel is a 3 × 3 matrix and the convolution step length is 1.
7. A prediction and early warning method for a lane change behavior of a vehicle is characterized by comprising the following steps:
step one, collecting vehicle driving state data:
a vehicle-mounted terminal is arranged in each vehicle, and the vehicle-mounted terminal integrates a CAN bus module to collect vehicle running data, a GPS module to acquire positioning data of the vehicles, and a communication module to realize data communication between the vehicles and a display module to display the collected vehicle running state data; the vehicle travel data includes a speed, an acceleration, and a steering angle of the vehicle;
step two, data communication between vehicles:
each vehicle transmits its vehicle driving state data to other vehicles through its communication module and receives the vehicle driving state data of the other vehicles;
step three, obtaining the lane change intention probability of the lane change behavior of the target vehicle by the method in the step 3 in the vehicle lane change behavior prediction model building method in claim 5;
step four, taking the driving state data of the target vehicle and the lane change intention probability as input, importing the data into the trained deep learning algorithm model of any claim from 1 to 6, and predicting whether the target vehicle changes lanes or not; and if the target vehicle is predicted to change the lane, displaying early warning information on vehicle-mounted terminals of other vehicles.
8. The utility model provides a vehicle lane change behavior prediction early warning system which characterized in that includes:
the vehicle running state data acquisition module comprises a vehicle-mounted terminal arranged in each vehicle, wherein the vehicle-mounted terminal integrates a CAN bus module to acquire vehicle running data, a GPS module to acquire positioning data of the vehicles, and a communication module to realize data communication between the vehicles and a display module to display the acquired vehicle running state data;
the inter-vehicle data communication module is used for transmitting the vehicle driving state data of each vehicle to other vehicles and receiving the vehicle driving state data of other vehicles;
the lane change intention probability acquisition module is used for establishing a game matrix by combining the driving state data of the vehicle after the target vehicle receives the driving state data of other vehicles and utilizing a game theory after the target vehicle receives the driving state data of the other vehicles, and then solving the game matrix to acquire the lane change intention probability of the target vehicle;
the target vehicle lane change prediction and early warning module is used for importing the driving state data of the target vehicle and the lane change intention probability into a trained deep learning algorithm model to predict whether the target vehicle changes lanes or not; and if the target vehicle is predicted to change the lane, displaying early warning information on vehicle-mounted terminals of other vehicles.
CN202110702854.4A 2021-06-24 2021-06-24 Vehicle lane change behavior prediction model construction, prediction and early warning method and system Active CN113516846B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110702854.4A CN113516846B (en) 2021-06-24 2021-06-24 Vehicle lane change behavior prediction model construction, prediction and early warning method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110702854.4A CN113516846B (en) 2021-06-24 2021-06-24 Vehicle lane change behavior prediction model construction, prediction and early warning method and system

Publications (2)

Publication Number Publication Date
CN113516846A true CN113516846A (en) 2021-10-19
CN113516846B CN113516846B (en) 2022-12-13

Family

ID=78066230

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110702854.4A Active CN113516846B (en) 2021-06-24 2021-06-24 Vehicle lane change behavior prediction model construction, prediction and early warning method and system

Country Status (1)

Country Link
CN (1) CN113516846B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113997954A (en) * 2021-11-29 2022-02-01 广州文远知行科技有限公司 Vehicle driving intention prediction method, device and equipment and readable storage medium
CN114162144A (en) * 2022-01-06 2022-03-11 苏州挚途科技有限公司 Automatic driving decision method and device and electronic equipment
CN115412883A (en) * 2022-08-31 2022-11-29 重庆交通大学 Intelligent network connection over-the-horizon driving auxiliary system based on 5G position sharing
CN115909749A (en) * 2023-01-09 2023-04-04 广州通达汽车电气股份有限公司 Vehicle operation road risk early warning method, device, equipment and storage medium

Citations (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2006087282A1 (en) * 2005-02-21 2006-08-24 Robert Bosch Gmbh Method for recognising an imminent overtake
EP2098431A2 (en) * 2008-02-27 2009-09-09 Audi AG Motor vehicle with driver assistance systems
CN108995655A (en) * 2018-07-06 2018-12-14 北京理工大学 A kind of driver's driving intention recognition methods and system
CN109460023A (en) * 2018-11-09 2019-03-12 上海理工大学 Driver's lane-changing intention recognition methods based on Hidden Markov Model
CN110097785A (en) * 2019-05-30 2019-08-06 长安大学 A kind of front truck incision or urgent lane-change identification prior-warning device and method for early warning
US20190344714A1 (en) * 2018-05-09 2019-11-14 Denso Corporation Lane changer warning system and method of the same
CN111081065A (en) * 2019-12-13 2020-04-28 北京理工大学 Intelligent vehicle collaborative lane change decision model under road section mixed traveling condition
CN111267846A (en) * 2020-02-11 2020-06-12 南京航空航天大学 Game theory-based peripheral vehicle interaction behavior prediction method
CN111275972A (en) * 2020-02-22 2020-06-12 长安大学 Automobile lane change early warning method and device based on V2V
CN111994079A (en) * 2020-09-18 2020-11-27 南京航空航天大学 Non-cooperative game lane change auxiliary decision making system and method considering driving style characteristics
CN111994089A (en) * 2020-09-02 2020-11-27 中国科学技术大学 Driver lane change intention identification method and system based on hybrid strategy game
CN112116100A (en) * 2020-09-08 2020-12-22 南京航空航天大学 Game theory decision method considering driver types
CN112163525A (en) * 2020-09-29 2021-01-01 新华三信息安全技术有限公司 Event type prediction method and device, electronic equipment and storage medium
CN112382115A (en) * 2020-10-29 2021-02-19 杭州电子科技大学 Driving risk early warning device and method based on visual perception
CN112562328A (en) * 2020-11-27 2021-03-26 腾讯科技（深圳）有限公司 Vehicle behavior prediction method and device
CN112614373A (en) * 2020-12-29 2021-04-06 厦门大学 BiLSTM-based weekly vehicle lane change intention prediction method
CN112789481A (en) * 2018-10-04 2021-05-11 祖克斯有限公司 Trajectory prediction for top-down scenarios

Patent Citations (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2006087282A1 (en) * 2005-02-21 2006-08-24 Robert Bosch Gmbh Method for recognising an imminent overtake
EP2098431A2 (en) * 2008-02-27 2009-09-09 Audi AG Motor vehicle with driver assistance systems
US20190344714A1 (en) * 2018-05-09 2019-11-14 Denso Corporation Lane changer warning system and method of the same
CN108995655A (en) * 2018-07-06 2018-12-14 北京理工大学 A kind of driver's driving intention recognition methods and system
CN112789481A (en) * 2018-10-04 2021-05-11 祖克斯有限公司 Trajectory prediction for top-down scenarios
CN109460023A (en) * 2018-11-09 2019-03-12 上海理工大学 Driver's lane-changing intention recognition methods based on Hidden Markov Model
CN110097785A (en) * 2019-05-30 2019-08-06 长安大学 A kind of front truck incision or urgent lane-change identification prior-warning device and method for early warning
CN111081065A (en) * 2019-12-13 2020-04-28 北京理工大学 Intelligent vehicle collaborative lane change decision model under road section mixed traveling condition
CN111267846A (en) * 2020-02-11 2020-06-12 南京航空航天大学 Game theory-based peripheral vehicle interaction behavior prediction method
CN111275972A (en) * 2020-02-22 2020-06-12 长安大学 Automobile lane change early warning method and device based on V2V
CN111994089A (en) * 2020-09-02 2020-11-27 中国科学技术大学 Driver lane change intention identification method and system based on hybrid strategy game
CN112116100A (en) * 2020-09-08 2020-12-22 南京航空航天大学 Game theory decision method considering driver types
CN111994079A (en) * 2020-09-18 2020-11-27 南京航空航天大学 Non-cooperative game lane change auxiliary decision making system and method considering driving style characteristics
CN112163525A (en) * 2020-09-29 2021-01-01 新华三信息安全技术有限公司 Event type prediction method and device, electronic equipment and storage medium
CN112382115A (en) * 2020-10-29 2021-02-19 杭州电子科技大学 Driving risk early warning device and method based on visual perception
CN112562328A (en) * 2020-11-27 2021-03-26 腾讯科技（深圳）有限公司 Vehicle behavior prediction method and device
CN112614373A (en) * 2020-12-29 2021-04-06 厦门大学 BiLSTM-based weekly vehicle lane change intention prediction method

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
EMMANUEL KOFI ADANU ET AL.: "Injury-severity analysis of lane change crashes involving commercial motor vehicles on interstate highways", 《EL SEVIER》 *
KANGQIANG OUYANG ET AL.: "Lane change decision planning for autonomous vehicles", 《IEEE》 *
宋晓琳 等: "基于模仿学习和强化学习的智能车辆换道行为决策", 《汽车工程》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113997954A (en) * 2021-11-29 2022-02-01 广州文远知行科技有限公司 Vehicle driving intention prediction method, device and equipment and readable storage medium
CN113997954B (en) * 2021-11-29 2023-11-21 广州文远知行科技有限公司 Method, device and equipment for predicting vehicle driving intention and readable storage medium
CN114162144A (en) * 2022-01-06 2022-03-11 苏州挚途科技有限公司 Automatic driving decision method and device and electronic equipment
CN114162144B (en) * 2022-01-06 2024-02-02 苏州挚途科技有限公司 Automatic driving decision method and device and electronic equipment
CN115412883A (en) * 2022-08-31 2022-11-29 重庆交通大学 Intelligent network connection over-the-horizon driving auxiliary system based on 5G position sharing
CN115909749A (en) * 2023-01-09 2023-04-04 广州通达汽车电气股份有限公司 Vehicle operation road risk early warning method, device, equipment and storage medium

Also Published As

Publication number Publication date
CN113516846B (en) 2022-12-13

Similar Documents

Publication Publication Date Title
CN113516846B (en) Vehicle lane change behavior prediction model construction, prediction and early warning method and system
Tang et al. Lane-changes prediction based on adaptive fuzzy neural network
CN107161153B (en) A kind of driving behavior methods of marking and device
Wang et al. How much data are enough? A statistical approach with case study on longitudinal driving behavior
CN112133089B (en) Vehicle track prediction method, system and device based on surrounding environment and behavior intention
Lyu et al. Vehicle trajectory prediction and cut-in collision warning model in a connected vehicle environment
US11465650B2 (en) Model-free reinforcement learning
US11480963B2 (en) Vehicle intent prediction neural network
DE102020103509A1 (en) DETECTING AND AVOIDING COLLISION BEHAVIOR
CN111145552B (en) Planning method for vehicle dynamic lane changing track based on 5G network
Deng et al. Modeling the effect of limited sight distance through fog on car-following performance using QN-ACTR cognitive architecture
CN112487954B (en) Pedestrian crossing behavior prediction method for plane intersection
Amsalu et al. Driver behavior modeling near intersections using hidden Markov model based on genetic algorithm
CN114655227A (en) Driving style recognition method, driving assistance method and device
Kim et al. Toward explainable and advisable model for self‐driving cars
Xu et al. The hybrid model for lane-changing detection at freeway off-ramps using naturalistic driving trajectories
Bian et al. Influence of prompt timing and messages of an audio navigation system on driver behavior on an urban expressway with five exits
Ontanon et al. Learning to predict driver behavior from observation
Wang et al. A state dependent mandatory lane-changing model for urban arterials with hidden Markov model method
KR20220089772A (en) Drive ability evaluated method for evaluating driving ability of driver and, human factor evaluation system for performing the method
Gao et al. A personalized model for driver lane-changing behavior prediction using deep neural network
CN116842340A (en) Pedestrian track prediction method and equipment based on ATT-GRU model
Lv et al. Traffic safety detection system by digital twins and virtual reality technology
CN116501820A (en) Vehicle track prediction method, device, equipment and storage medium
CN115771506A (en) Method and device for determining vehicle driving strategy based on passenger risk cognition

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant