CN112258893A - Intelligent vehicle lane change collision probability assessment method based on track prediction - Google Patents

Intelligent vehicle lane change collision probability assessment method based on track prediction Download PDF

Info

Publication number
CN112258893A
CN112258893A CN202010902992.2A CN202010902992A CN112258893A CN 112258893 A CN112258893 A CN 112258893A CN 202010902992 A CN202010902992 A CN 202010902992A CN 112258893 A CN112258893 A CN 112258893A
Authority
CN
China
Prior art keywords
vehicle
state
probability
time
hidden
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
CN202010902992.2A
Other languages
Chinese (zh)
Other versions
CN112258893B (en
Inventor
段雨宸
邓昱聪
漆巍巍
钟浩川
吴丽莎
谭永鑫
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.)
South China University of Technology SCUT
Original Assignee
South China University of Technology SCUT
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 South China University of Technology SCUT filed Critical South China University of Technology SCUT
Priority to CN202010902992.2A priority Critical patent/CN112258893B/en
Publication of CN112258893A publication Critical patent/CN112258893A/en
Application granted granted Critical
Publication of CN112258893B publication Critical patent/CN112258893B/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
    • G08G1/16—Anti-collision systems
    • G08G1/164—Centralised systems, e.g. external to vehicles
    • 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
    • B60W30/08—Active safety systems predicting or avoiding probable or impending collision or attempting to minimise its consequences
    • B60W30/095—Predicting travel path or likelihood of collision

Landscapes

  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Transportation (AREA)
  • Mechanical Engineering (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Traffic Control Systems (AREA)

Abstract

The invention discloses an intelligent vehicle lane change collision probability assessment method based on track prediction, which comprises the following steps: 1) obtaining the size of a vehicle, real-time motion parameters of the vehicle and longitudinal relative distance between a target vehicle and surrounding vehicles; 2) establishing an original coordinate system of collision prediction by taking the centroid position of the target vehicle as an origin and the centroid position of the vehicle as the approximate position of the vehicle, and carrying out grid division on the coordinate system; 3) predicting the probability of the target vehicle reaching each grid by using a hidden Markov model; 4) analyzing according to the result output by the forward algorithm and calculating the relative displacement of the target vehicle and the surrounding vehicles after the time t; 5) calculating the relative positions of the target vehicle and the surrounding vehicles after t time according to the initial positions, the motion states and the prediction information of the vehicles; 6) and (4) visually outputting the probability distribution of the target vehicle and the surrounding vehicles reaching each grid at the time t by using a thermodynamic diagram. The method can effectively solve the problem of collision probability prediction possibly existing in midway lane changing of the vehicle.

Description

Intelligent vehicle lane change collision probability assessment method based on track prediction
Technical Field
The invention relates to the technical field of safety and evaluation of lane changing of motor vehicles on urban roads and expressways, in particular to an intelligent vehicle lane changing collision probability evaluation method based on track prediction.
Background
When the motor vehicle runs on the road, because the front vehicle speed is slower and the rear vehicle speed is faster, the overtaking driving behavior is generated in order to pursue higher time benefit and space resource. The specific ways for realizing the overtaking behaviors are various, and overtaking is generally implemented on a front vehicle by changing lanes of a rear vehicle. However, the behavior of changing lanes inevitably affects the traveling of surrounding vehicles, and if the operation of changing lanes is not performed under appropriate conditions, there is a possibility that the vehicle normally traveling behind collides with the lane, thereby causing a traffic accident. According to the statistics of the traffic accident data of the roads in the United states, the traffic accidents caused by bad lane changing behaviors account for about 5 percent of the total number of accidents. Therefore, it is necessary to consider the road safety factor when changing lanes and calculate the collision probability of the motor vehicle caused by lane change, and predict and grade the collision.
At present, domestic judgment and grading research on intelligent lane change collision is not enough, and similar research is more stopped at judging whether lane change actions can generate collision among moving vehicles, and the probability of collision is not realized in a quantification and early warning mode, so that collision judgment is still stopped in two simple states of collision and non-collision. This also results in the severity of the impact of the lane-change control action on the driving state of the vehicle, which cannot be clearly controlled by the driver when preparing to change lanes, and thus affects the driving behavior of the driver to a certain extent. The invention converts the conflict point problem generated by the vehicle lane changing motion into the potential collision area problem generated by the running of the entity vehicle, divides grids in a coordinate system taking the lane changing vehicle as an origin point in order to determine the potential collision area and the corresponding collision probability, calculates the vehicle collision probability possibly generated by any grid in the coordinate system on the premise of considering the time, the longitudinal speed of the front vehicle and the rear vehicle, the transverse displacement trend and other factors, and divides the conflict probability according to the set collision threshold value, thereby achieving the aim of early warning.
Disclosure of Invention
The invention aims to overcome the defects and shortcomings of the prior art, and provides an intelligent vehicle lane change collision probability assessment method based on track prediction.
In order to achieve the purpose, the technical scheme provided by the invention is as follows: an intelligent vehicle lane change collision probability assessment method based on track prediction comprises the following steps:
1) selecting a road section where a target vehicle is located, and acquiring the size of the vehicle, real-time motion parameters of the vehicle and the longitudinal relative distance L between the target vehicle and surrounding vehicles0；
2) Establishing an original coordinate system of conflict prediction, defining M, N the length and width of the minimum grid unit respectively, and uniformly dividing the original coordinate system into 2 Mx 3N grids;
3) and (3) calculating the collision prediction probability: predicting the probability of the target vehicle reaching each grid by using a Hidden Markov Model (HMM), inputting the current speed and acceleration of the lane changing vehicle, and predicting the position and the acceleration state of the target vehicle at the moment when the target vehicle changes lanes by using a forward algorithm;
4) analyzing according to the result output by the forward algorithm, and calculating the relative displacement D of the target vehicle and the surrounding vehicles after the time ti，Di＝(Li-L1) In the formula, DiIs the relative displacement of the target vehicle and the surrounding vehicle at the time point t，LiFor longitudinal displacement of the surrounding vehicle in the time interval 0-t, L1The longitudinal displacement of the target vehicle in a time interval of 0-t, wherein t is a prediction time interval;
5) judging lane changing collision conditions: according to the initial position, the motion state and the prediction information of each vehicle, calculating the relative position of the target vehicle and the surrounding vehicles after t time, and determining the collision critical position so as to deduce the collision probability PH，PHCalculating the sum of the predicted collision probabilities for each grid;
6) programming a forward algorithm by utilizing Matlab, visually outputting probability distribution of the target vehicle and surrounding vehicles reaching each grid at the time t by using a thermodynamic diagram, and defining PtIs a collision warning probability threshold, if PH<PtIf P is not early-warning, then P is not early-warningH>PtAnd the vehicle man-machine interaction system is used for early warning the driver so as to achieve the purpose of predicting lane change collision.
In step 1), the vehicle size includes a vehicle length and a width, and can be obtained from vehicle management registered data; the real-time motion parameters of the vehicle comprise the instantaneous speed and the instantaneous acceleration of the vehicle, and can be obtained through a motion sensor of the vehicle; the longitudinal relative distance L between the target vehicle and the surrounding vehicle0Can be obtained by a camera, a millimeter wave radar and an ultrasonic radar around the vehicle.
In the step 2), an original coordinate system of the collision prediction is established by taking the center of mass of the target vehicle as an origin and the position of the center of mass of the vehicle as the approximate position of the center of mass of the vehicle, and the coordinate system is uniformly divided into 2 Mx 3N grids, wherein the size of each grid is not larger than 1 standard vehicle size.
In step 3), according to the obtained original coordinate system of the target road section, the vehicle motion parameters and the longitudinal relative distance of the target vehicle and the surrounding vehicles, performing collision prediction probability calculation, and comprising the following steps:
3.1) the position and the speed of a vehicle running on a lane are sampled in a segmented manner, the position and the speed of the vehicle are changed in sampling time, meanwhile, the hidden state corresponding to the position of the vehicle is changed, and the vehicle is changed from the current position (observable state) to the destination position, wherein the process is defined as the process of vehicle observation state transition; in the HMM model, the speed and position of the vehicle are associated with different motion patterns, which are rapid acceleration, slow acceleration, steady-state driving, slow deceleration, and rapid deceleration, while in reality, the motion patterns are difficult to directly observe, so the motion patterns are used as hidden states of the HMM model;
in summary, an HMM model based on an observed state and an implied state is constructed, and an expression of the HMM model is as follows:
λ＝(A,B,π,n,m)
wherein the content of the first and second substances,
πi＝p(ti),1≤i≤n
Figure BDA0002660398880000041
wherein
Figure BDA0002660398880000044
Figure BDA0002660398880000042
Wherein B isjk＝p(k|tj)
In the formula, lambda represents a quintuple, and the HMM model is determined by a state transition probability matrix A, a confusion matrix B, an initial probability distribution pi, a set hidden state number n and an observation state number m in a certain hidden state, so that the HMM model is represented by the quintuple lambda; pi ═ pi1,π2,...,πi) Representing an initial probability distribution of the HMM model; p (t)i) Denotes the initial probability of occurrence of an implicit state i at a time when the system time T is 0, i denotes each implicit state, n implicit states in total, piiRepresenting an initial probability of occurrence of an implicit state i at a time when the system time T ═ 0; a represents a state transition matrix in the HMM model; in A matrix, tj TIndicating that the sampling time T is in an implicit state i, Tj T+1Represents the time of T +1In the implicit state j,
Figure BDA0002660398880000043
representing the probability of transition from an implicit state i at the sampling time T to an implicit state j at T +1, the element A in the matrixijRepresenting the transition probability from the hidden state i at the sampling time T to the hidden state j at the time T + 1; defining hidden states as different motion modes in the driving process, wherein B represents a confusion matrix in the HMM model, and an element B of the matrixjkRepresenting the probability of observing observation state k at the occurrence of hidden state j; each hidden state in the sequence corresponds to an observed state; the observation state corresponds to different movement modes, and the vehicle can reach different positions, namely grids, on the lane; n represents the number of the set hidden states, and the maximum value of the number of the set hidden states changes along with the number of the set hidden states during modeling; m represents the number of observation states of the hidden Markov algorithm in a specific hidden state;
3.2) determining the parameter value of the HMM model, and calculating the model parameter with the highest possibility by adopting an EM algorithm;
3.2.1) π -by- π in the Lagrangian function in EM algorithmsiAnd obtaining a partial derivative, namely:
Figure BDA0002660398880000051
in the formula: the observation sequence O (O) is given in advance between 0 and T1,o2，...,oT) The hidden sequence I ═ I (I)1,i2,...iT) Wherein o isTIndicating the observed state at each time, iTRepresenting the implicit state result at each time instant; piiRepresenting an initial probability of occurrence of an implicit state i at a time when the system time T ═ 0;
Figure BDA0002660398880000052
is a current estimate of the HMM parameter; gamma is the probability of a single hidden state;
Figure BDA0002660398880000053
indicates that any one hidden sequence I ═ is present given the HMM model estimate values (I ═1,i2,...iT) The probability of (d); p (O | I, λ) denotes for a fixed hidden state sequence I ═ I (I)1,i2,...iT) The appearance of the observation sequence O ═ O (O)1,o2，...,oT) Probability of, then
Figure BDA0002660398880000054
Representing the probability of the joint occurrence of O and I;
Figure BDA0002660398880000055
representing the conditional probability of the observation sequence O occurring after the HMM model parameter values are given;
3.2.2) A in A matrix elementsijIs calculated iteratively
Figure BDA0002660398880000056
In the formula (I), the compound is shown in the specification,
Figure BDA0002660398880000057
is the current estimated value of the HMM model parameter; the observation sequence O (O) is given in advance between 0 and T1,o2，...,oT) The hidden sequence I ═ I (I)1,i2,...iT) Wherein o isTIndicating the observed state at each time, iTRepresenting the implicit state result at each time instant; a. theijRepresenting the transition probability from the hidden state i at the sampling time T to the hidden state j at the time T + 1;
Figure BDA0002660398880000061
indicating that, given HMM model parameters, an observation sequence O ═ occurs when the hidden state at the previous time is i and the hidden state at the next time is j1,o2，...,oT) The probability of (d);
Figure BDA0002660398880000062
expressed at a given HMUnder the M model parameter, the hidden state is i, and the appearance observation sequence O ═ O1,o2，...,oT) The probability of (d);
3.2.3) B in B matrix elementsjkIs calculated iteratively
Figure BDA0002660398880000063
In the formula:
Figure BDA0002660398880000064
is the current estimated value of the HMM model parameter; the observation sequence O (O) is given in advance between 0 and T1,o2，...,oT) The hidden sequence I ═ I (I)1,i2,...iT) Wherein o isTIndicating the observed state at each time, iTRepresenting the implicit state result at each time instant;
Figure BDA0002660398880000065
indicating that, given HMM model parameters, the hidden state is i, and the observation sequence O ═ is present (O)1,o2，...,oT) The probability of (d); i (o)nK) is a decision factor, if and only if the observed state onK is 1, otherwise 0;
through the above calculation, the HMM parameter λ of the current automobile can be obtained as (pi, a, B, n, m).
In step 5), calculating the relative positions of the target vehicle and the surrounding vehicles after t time according to the initial positions, the motion states and the prediction information of the vehicles, and determining collision critical positions so as to deduce the collision probability;
Figure BDA0002660398880000066
PH＝∑Pij
in the formula, DiIs the relative displacement between the target vehicle and the surrounding vehicle at time t, L0Indicating initial longitudinal position of target and surrounding vehiclesShift difference, H represents vehicle collision state, when H is 0, Di>L0When the target vehicle collides with the surrounding vehicle; when H is 1, i.e. Di<L0When the target vehicle does not collide with the surrounding vehicle; pijThe predicted collision probability is calculated for each grid, and the relative displacement between the target vehicle and the surrounding vehicle is D after t timeiAnd Di>L0The probability of (c).
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. the invention can carry out safety evaluation on the lane changing action, and further carry out early warning on a driver through a human-computer interaction system under the background of the intelligent vehicle, thereby reducing the accident occurrence probability.
2. According to the invention, the vehicle is not abstracted into mass points in a general manner, but the size and the microscopic running track of the vehicle, the motion parameters and the probability distribution reaching each grid are considered, so that the actual running condition of the vehicle on the road is better met, and the prediction result is more accurate.
3. The invention further improves the prediction precision of the motor vehicle lane change prediction, does not only solve the problem of simplicity but also expresses the problem by using specific probability, and has higher applicability and application universality compared with the prior research.
Drawings
FIG. 1 is a schematic diagram of the road grid division where a target vehicle is located according to the present invention.
Fig. 2 is a diagram showing a predicted position distribution of a surrounding vehicle after time t in the present invention.
Fig. 3 is a diagram showing a relative positional relationship between the target vehicle and the nearby vehicle in the present invention.
Detailed Description
The present invention will be further described with reference to the following specific examples.
The intelligent vehicle lane change collision probability assessment method based on track prediction provided by the embodiment comprises the following steps:
1) selecting the road section where the target vehicle is located, and acquiring the size of the vehicle, the real-time motion parameters of the vehicle, the target vehicle and the surrounding vehiclesLongitudinal relative distance L0。
The vehicle size includes a vehicle length and a vehicle width, and can be obtained from data registered in a vehicle management; the real-time motion parameters of the vehicle comprise the instantaneous speed and the instantaneous acceleration of the vehicle, and can be obtained through a motion sensor of the vehicle; the longitudinal relative distance L between the target vehicle and the surrounding vehicle0Can be obtained by a camera, a millimeter wave radar and an ultrasonic radar around the vehicle.
2) And establishing an original coordinate system of the collision prediction by taking the position of the center of mass of the target vehicle as an origin and the position of the center of mass of the vehicle as the approximate position of the center of mass of the vehicle, and defining M, N the length and the width of the minimum grid unit respectively. The coordinate system is evenly divided into 2 Mx 3N grids, the size of each grid is not larger than that of 1 standard vehicle, and coordinates of the target vehicle and surrounding vehicles are calibrated.
3) And (3) calculating the collision prediction probability: and predicting the probability of the target vehicle reaching each grid by using a Hidden Markov Model (HMM), inputting the current speed and acceleration of the lane changing vehicle, and predicting the position and the acceleration state of the target vehicle at the moment when the target vehicle changes lanes by using a forward algorithm.
According to the obtained original coordinate system of the target road section, the vehicle motion parameters and the longitudinal relative distance of the target vehicle and the surrounding vehicles, the collision prediction probability calculation is carried out, and the method comprises the following steps:
3.1) the position and the speed of a vehicle running on a lane are sampled in a segmented manner, the position and the speed of the vehicle are changed in sampling time, meanwhile, the hidden state corresponding to the position of the vehicle is changed, and the vehicle is changed from the current position (observable state) to the destination position, wherein the process is defined as the process of vehicle observation state transition; in the HMM model, the speed and position of the vehicle are associated with different motion patterns, which are rapid acceleration, slow acceleration, steady-state driving, slow deceleration, and rapid deceleration, while in reality, the motion patterns are difficult to directly observe, so the motion patterns are used as hidden states of the HMM model;
in summary, an HMM model based on an observed state and an implied state is constructed, and an expression of the HMM model is as follows:
λ＝(A,B,π,n,m)
wherein the content of the first and second substances,
πi＝p(ti),1≤i≤n
Figure BDA0002660398880000091
wherein
Figure BDA0002660398880000092
Figure BDA0002660398880000093
Wherein B isjk＝p(k|tj)
In the invention, an HMM model is determined by a state transition probability matrix A, a confusion matrix B, an initial probability distribution pi, a set hidden state number n and an observation state number m under a certain hidden state, so that the HMM model is expressed by a quintuple lambda; pi ═ pi1,π2,...,πi) Representing an initial probability distribution of the HMM model; p (t)i) Denotes the initial probability of occurrence of an implicit state i at a time when the system time T is 0, i denotes each implicit state, n implicit states in total, piiRepresenting an initial probability of occurrence of an implicit state i at a time when the system time T ═ 0; a represents a state transition matrix in the HMM model; in A matrix, tj TIndicating that the sampling time T is in an implicit state i, Tj T+1Indicating that time T +1 is in implicit state j,
Figure BDA0002660398880000094
representing the probability of transition from an implicit state i at the sampling time T to an implicit state j at T +1, the element A in the matrixijRepresenting the transition probability from the hidden state i at the sampling time T to the hidden state j at the time T + 1; defining hidden states as different motion modes in the driving process, wherein B represents a confusion matrix in an HMM model; elements of the matrixBjkRepresenting the probability of observing observation state k at the occurrence of hidden state j; each hidden state in the sequence corresponds to an observed state; the observation state corresponds to different movement modes, and the vehicle can reach different positions, namely grids, on the lane; n represents the number of the set hidden states, and the maximum value of the number of the set hidden states changes along with the number of the set hidden states during modeling; m represents the number of observation states of the hidden Markov algorithm in a specific hidden state;
3.2) determining the parameter value of the HMM model, and calculating the model parameter with the highest possibility by adopting an EM algorithm;
3.2.1) π -by- π in the Lagrangian function in EM algorithmsiAnd obtaining a partial derivative, namely:
Figure BDA0002660398880000101
in the formula: the observation sequence O (O) is given in advance between 0 and T1,o2，...,oT) The hidden sequence I ═ I (I)1,i2,...iT) Wherein o isTIndicating the observed state at each time, iTRepresenting the implicit state result at each time instant; piiRepresenting an initial probability of occurrence of an implicit state i at a time when the system time T ═ 0;
Figure BDA0002660398880000102
is a current estimate of the HMM parameter; gamma is the probability of a single hidden state;
Figure BDA0002660398880000103
indicates that any one hidden sequence I ═ is present given the HMM model estimate values (I ═1,i2,...iT) The probability of (d); p (O | I, λ) denotes for a fixed hidden state sequence I ═ I (I)1,i2,...iT) The appearance of the observation sequence O ═ O (O)1,o2，...,oT) Probability of, then
Figure BDA0002660398880000104
Representing the probability of the joint occurrence of O and I;
Figure BDA0002660398880000105
representing the conditional probability of the observation sequence O occurring after the HMM model parameter values are given;
3.2.2) A in A matrix elementsijIs calculated iteratively
Figure BDA0002660398880000106
In the formula (I), the compound is shown in the specification,
Figure BDA0002660398880000107
is the current estimated value of the HMM model parameter; the observation sequence O (O) is given in advance between 0 and T1,o2，...,oT) The hidden sequence I ═ I (I)1,i2,...iT) Wherein o isTIndicating the observed state at each time, iTRepresenting the implicit state result at each time instant; a. theijRepresenting the transition probability from the hidden state i at the sampling time T to the hidden state j at the time T + 1;
Figure BDA0002660398880000108
indicating that, given HMM model parameters, an observation sequence O ═ occurs when the hidden state at the previous time is i and the hidden state at the next time is j1,o2，...,oT) The probability of (d);
Figure BDA0002660398880000109
indicating that, given HMM model parameters, the hidden state is i, and the observation sequence O ═ is present (O)1,o2，...,oT) The probability of (d);
3.2.3) B in B matrix elementsjkIs calculated iteratively
Figure BDA0002660398880000111
In the formula:
Figure BDA0002660398880000112
is the current estimated value of the HMM model parameter; the observation sequence O (O) is given in advance between 0 and T1,o2，...,oT) The hidden sequence I ═ I (I)1,i2,...iT) Wherein o isTIndicating the observed state at each time, iTRepresenting the implicit state result at each time instant;
Figure BDA0002660398880000113
indicating that, given HMM model parameters, the hidden state is i, and the observation sequence O ═ is present (O)1,o2，...,oT) The probability of (d); i (o)nK) is a decision factor, if and only if the observed state onK is 1, otherwise 0.
Through the above calculation, the HMM model parameter λ of the current automobile can be obtained as (pi, a, B, n, m).
4) Calculating the relative displacement D of the target vehicle and the surrounding vehicles after the time ti，Di＝(Li-L1) In the formula, DiIs the relative displacement between the target vehicle and the surrounding vehicle at time t, LiFor longitudinal displacement of the surrounding vehicle in the time interval 0-t, L1The longitudinal displacement of the target vehicle in a time period from 0 to t is obtained, and the time t is a prediction time interval.
5) Judging lane changing collision conditions: according to the initial position, the motion state and the prediction information of each vehicle, calculating the relative position of the target vehicle and the surrounding vehicles after t time, and determining the collision critical position so as to deduce the collision probability PH；
Figure BDA0002660398880000114
PH＝∑Pij
In the formula, DiIs the relative displacement between the target vehicle and the surrounding vehicle at time t, L0Indicating the initial longitudinal displacement difference between the target vehicle and the surrounding vehicle, and H indicating the vehicle collision stateWhen H is 0, i.e. Di>L0When the target vehicle collides with the surrounding vehicle; when H is 1, i.e. Di<L0When the target vehicle does not collide with the surrounding vehicle; pijThe predicted collision probability is calculated for each grid, and the relative displacement between the target vehicle and the surrounding vehicle is D after t timeiAnd Di>L0Probability of (P)HThe sum of the resulting predicted collision probabilities is calculated for each grid.
6) Programming a forward algorithm by utilizing Matlab, visually outputting probability distribution of the target vehicle and surrounding vehicles reaching each grid at the time t by using a thermodynamic diagram, and defining PtIs a collision warning probability threshold, if PH<PtIf P is not early-warning, then P is not early-warningH>PtAnd the vehicle man-machine interaction system is used for early warning the driver so as to achieve the purpose of predicting lane change collision.
The existing observation road section is 70m long and 9m wide, and has 3 lanes, a series of observed vehicle motion parameter data exist, one vehicle is selected as a target vehicle A, and the collision probability of lane change of the vehicle after 1.5s is predicted by applying the method. It is known that there are B vehicles at 4m behind the left lane of the vehicle, and there are five driving behavior possibilities after t is 1.5s for a vehicle a, namely, rapid acceleration, slow acceleration, steady running, slow deceleration, and rapid deceleration. And the equal probability of each driving behavior of the A vehicle is set. Known conflict warning threshold PtThe vehicle initial position is shown in fig. 1 at 5%.
1) And acquiring required data. The observed data is known to be recorded by a camera of an urban road, the width of a lane is 3.75 meters, and the speed of a vehicle B is 15.23 m/s. The vehicle speed of the A vehicle is 13m/s, and the acceleration is-3.31 m/s2. The vehicle A and the vehicle B can be similar to standard vehicles, and the length is 3.5 m.
2) Taking the vehicle a as a coordinate origin, assuming that M is 10, N is 1, and 2M × 3N is 60, that is, a road surface with a length of 70M and a width of 9M is divided into 60 grids, each grid is 3.5M long and 3M wide. Wherein xjIs the abscissa, y, of the j-th column of pointsiIs the ordinate of the ith row of division points.
3) And establishing a model according to a hidden Markov algorithm. There is a possibility that five driving behaviors of the vehicle ahead exist, that initial probability distribution can be set as (0.2,0.2,0.2,0.2,0.2, 0.2), and we get a set of states as:
q ═ 5 ═ rapid acceleration, slow acceleration, steady-state travel, slow deceleration, and rapid deceleration }, N ═ slow acceleration, slow deceleration, and rapid deceleration
The observation set is as follows:
V＝{L1，L2...L5}，M＝5
3.1) we assume that each driving behavior has a certain possibility to change at the next observation time, that is, the probability of transition between hidden states, for example, the probability that the next observation time of each driving behavior still remains unchanged can be assumed to be 0.5, the probability of converting from rapid acceleration to rapid acceleration is 0.25, the probability of converting into steady-state driving is 0.15, and the probability of converting into rapid deceleration and slow deceleration is 0.05, and the state transition matrix can be obtained as follows in turn:
Figure BDA0002660398880000131
3.2) the probability of reaching different positions in the same driving time is different for each driving behavior, and we assume five different positions in the previously established coordinate system, as shown in FIG. 2:
the probability that different driving behaviors will reach different positions will also change. Assuming a rapid acceleration, L is reachediThe probabilities of (a) are 0.05, 0.1, 0.25, 0.55, respectively. In the same way, the probability of the slow acceleration, the steady-state running, the slow deceleration and the rapid deceleration reaching Li can be obtained.
The probability matrix of the observed state is obtained as follows:
Figure BDA0002660398880000132
3.3) starting the prediction of the position of the vehicle at the next instant after a, B, pi has been given. Assuming that we choose the first moment (as shown in fig. 1) where the speed is 13.71m/s and the acceleration is 0.01m/s, we can consider steady-state driving (different definitions and parameters for driving behavior such as rapid acceleration, rapid deceleration, steady-state driving, etc. under different circumstances).
3.4) when the steady state driving is carried out at the time when t is equal to 0, the probability that the position 1 is reached at the next time is obtained by using the state transition matrix and the observation state matrix:
PB1＝0.05*0.5+0.2*0.05+0.5*0.1+0.2*0.2+0.05*0.6＝0.1325
the same can get:
PB2＝0.05*0.05+0.2*0.05+0.5*0.2+0.2*0.55+0.05*0.2＝0.2325
PB3＝0.05*0.1+0.2*0.15+0.5*0.5+0.2*0.15+0.05*0.1＝0.3200
PB4＝0.05*0.25+0.2*0.55+0.5*0.2+0.2*0.05+0.05*0.05＝0.2350
PB5＝0.05*0.55+0.2*0.2+0.5*0.1+0.2*0.05+0.05*0.05＝0.1300
4) for the A car, the speed is 10m/s at this time, and the speed difference from the next time is 1.63m/s2Also known as rapid acceleration, the probability of the vehicle reaching each location is obtained using the above probability calculation:
PA1＝0.5*0.05+0.25*0.05+0.15*0.1+0.05*0.2+0.05*0.6＝0.0925
PA2＝0.5*0.05+0.25*0.05+0.15*0.2+0.05*0.55+0.05*0.2＝0.1050
PA3＝0.5*0.1+0.25*0.15+0.15*0.5+0.05*0.15+0.05*0.1＝0.1750
PA4＝0.5*0.25+0.25*0.55+0.15*0.2+0.05*0.05+0.05*0.05＝0.2975
PA5＝0.5*0.55+0.25*0.2+0.15*0.1+0.05*0.05+0.05*0.05＝0.3450
5) since A, B cars have a longitudinal headway of 4m initially, a collision situation may only occur at L which B cars may arrive in the future when a car a changes lanes after 1.5s3、L4、L5I.e. L that the A car may reach1、L2、L3As shown in FIG. 2 and the drawings3, respectively.
6) In summary, when the vehicle a and the vehicle B approximately coincide in the longitudinal direction, there is a collision risk when lane changing is performed, and the specific collision probability is calculated as follows:
PH1＝PB3×PA1＝0.3200×0.0925＝0.0296
PH2＝PB4×PA2＝0.2350×0.1050＝0.024675
PH3＝PB5×PA3＝0.1300×0.1750＝0.02275
PH＝PH1+PH2+PH3＝0.05565≈5.6％
known conflict warning threshold Pt＝5％，PH>PtTherefore, the driver should be sent an early warning signal, and at the moment, the lane changing is carried out, so that certain safety risk is caused.
The above-mentioned embodiments are merely preferred embodiments of the present invention, and the scope of the present invention is not limited thereto, so that the changes in the shape and principle of the present invention should be covered within the protection scope of the present invention.

Claims (5)

1. An intelligent vehicle lane change collision probability assessment method based on track prediction is characterized by comprising the following steps:
1) selecting a road section where a target vehicle is located, and acquiring the size of the vehicle, real-time motion parameters of the vehicle and the longitudinal relative distance L between the target vehicle and surrounding vehicles0；
2) Establishing an original coordinate system of conflict prediction, defining M, N the length and width of the minimum grid unit respectively, and uniformly dividing the original coordinate system into 2 Mx 3N grids;
3) and (3) calculating the collision prediction probability: predicting the probability of the target vehicle reaching each grid by using a Hidden Markov Model (HMM), inputting the current speed and acceleration of the lane changing vehicle, and predicting the position and the acceleration state of the target vehicle at the moment when the target vehicle changes lanes by using a forward algorithm;
4) analyzing according to the result output by the forward algorithmCalculating the relative displacement D between the target vehicle and the surrounding vehicles after the time ti，Di＝(Li-L1) In the formula, DiIs the relative displacement between the target vehicle and the surrounding vehicle at time t, LiFor longitudinal displacement of the surrounding vehicle in the time interval 0-t, L1The longitudinal displacement of the target vehicle in a time interval of 0-t, wherein t is a prediction time interval;
5) judging lane changing collision conditions: according to the initial position, the motion state and the prediction information of each vehicle, calculating the relative position of the target vehicle and the surrounding vehicles after t time, and determining the collision critical position so as to deduce the collision probability PH，PHCalculating the sum of the predicted collision probabilities for each grid;
6) programming a forward algorithm by utilizing Matlab, visually outputting probability distribution of the target vehicle and surrounding vehicles reaching each grid at the time t by using a thermodynamic diagram, and defining PtIs a collision warning probability threshold, if PH<PtIf P is not early-warning, then P is not early-warningH>PtAnd the vehicle man-machine interaction system is used for early warning the driver so as to achieve the purpose of predicting lane change collision.
2. The intelligent vehicle lane change collision probability assessment method based on track prediction as claimed in claim 1, characterized in that: in step 1), the vehicle size includes a vehicle length and a width, and can be obtained from vehicle management registered data; the real-time motion parameters of the vehicle comprise the instantaneous speed and the instantaneous acceleration of the vehicle, and can be obtained through a motion sensor of the vehicle; the longitudinal relative distance L between the target vehicle and the surrounding vehicle0Can be obtained by a camera, a millimeter wave radar and an ultrasonic radar around the vehicle.
3. The intelligent vehicle lane change collision probability assessment method based on track prediction as claimed in claim 1, characterized in that: in the step 2), an original coordinate system of the collision prediction is established by taking the center of mass of the target vehicle as an origin and the position of the center of mass of the vehicle as the approximate position of the center of mass of the vehicle, and the coordinate system is uniformly divided into 2 Mx 3N grids, wherein the size of each grid is not larger than 1 standard vehicle size.
4. The intelligent vehicle lane change collision probability assessment method based on track prediction as claimed in claim 1, characterized in that: in step 3), according to the obtained original coordinate system of the target road section, the vehicle motion parameters and the longitudinal relative distance of the target vehicle and the surrounding vehicles, performing collision prediction probability calculation, and comprising the following steps:
3.1) sampling the position and the speed of a vehicle running on a lane in a segmented manner, wherein the position and the speed of the vehicle change within sampling time, meanwhile, the hidden state corresponding to the position of the vehicle also changes, and the vehicle changes from the current position to the destination position, and the process is defined as the process of the state transition of the vehicle observation; in the HMM model, the speed and position of the vehicle are associated with different motion patterns, which are rapid acceleration, slow acceleration, steady-state driving, slow deceleration, and rapid deceleration, while in reality, the motion patterns are difficult to directly observe, so the motion patterns are used as hidden states of the HMM model;
in summary, an HMM model based on an observed state and an implied state is constructed, and an expression of the HMM model is as follows:
λ＝(A,B,π,n,m)
wherein the content of the first and second substances,
πi＝p(ti),1≤i≤n
Figure FDA0002660398870000031
wherein
Figure FDA0002660398870000032
Figure FDA0002660398870000033
Wherein B isjk＝p(k|tj)
In the formula, lambda represents a quintuple, HMM modelThe method is determined by a state transition probability matrix A, a confusion matrix B, an initial probability distribution pi, a set hidden state number n and an observation state number m under a certain hidden state, so that an HMM model is expressed by a quintuple lambda; pi ═ pi1,π2,...,πi) Representing an initial probability distribution of the HMM model; p (t)i) Denotes the initial probability of occurrence of an implicit state i at a time when the system time T is 0, i denotes each implicit state, n implicit states in total, piiRepresenting an initial probability of occurrence of an implicit state i at a time when the system time T ═ 0; a represents a state transition matrix in the HMM model; in A matrix, tj TIndicating that the sampling time T is in an implicit state i, Tj T+1Indicating that time T +1 is in implicit state j,
Figure FDA0002660398870000034
representing the probability of transition from an implicit state i at the sampling time T to an implicit state j at T +1, the element A in the matrixijRepresenting the transition probability from the hidden state i at the sampling time T to the hidden state j at the time T + 1; defining hidden states as different motion modes in the driving process, wherein B represents a confusion matrix in the HMM model, and an element B of the matrixjkRepresenting the probability of observing observation state k at the occurrence of hidden state j; each hidden state in the sequence corresponds to an observed state; the observation state corresponds to different movement modes, and the vehicle can reach different positions, namely grids, on the lane; n represents the number of the set hidden states, and the maximum value of the number of the set hidden states changes along with the number of the set hidden states during modeling; m represents the number of observation states of the hidden Markov algorithm in a specific hidden state;
3.2) determining the parameter value of the HMM model, and calculating the model parameter with the highest possibility by adopting an EM algorithm;
3.2.1) π -by- π in the Lagrangian function in EM algorithmsiAnd obtaining a partial derivative, namely:
Figure FDA0002660398870000041
in the formula: the observation sequence O (O) is given in advance between 0 and T1,o2，...,oT) The hidden sequence I ═ I (I)1,i2,...iT) Wherein o isTIndicating the observed state at each time, iTRepresenting the implicit state result at each time instant; piiRepresenting an initial probability of occurrence of an implicit state i at a time when the system time T ═ 0;
Figure FDA0002660398870000042
is a current estimate of the HMM parameter; gamma is the probability of a single hidden state;
Figure FDA0002660398870000043
indicates that any one hidden sequence I ═ is present given the HMM model estimate values (I ═1,i2,...iT) The probability of (d); p (O | I, λ) denotes for a fixed hidden state sequence I ═ I (I)1,i2,...iT) The appearance of the observation sequence O ═ O (O)1,o2，...,oT) Probability of, then
Figure FDA0002660398870000044
Representing the probability of the joint occurrence of O and I;
Figure FDA0002660398870000045
representing the conditional probability of the observation sequence O occurring after the HMM model parameter values are given;
3.2.2) A in A matrix elementsijIs calculated iteratively
Figure FDA0002660398870000046
In the formula (I), the compound is shown in the specification,
Figure FDA0002660398870000047
is the current estimated value of the HMM model parameter; given in advance at 0 to TInscribed observation sequence O ═ O (O)1,o2，...,oT) The hidden sequence I ═ I (I)1,i2,...iT) Wherein o isTIndicating the observed state at each time, iTRepresenting the implicit state result at each time instant; a. theijRepresenting the transition probability from the hidden state i at the sampling time T to the hidden state j at the time T + 1;
Figure FDA0002660398870000048
indicating that, given HMM model parameters, an observation sequence O ═ occurs when the hidden state at the previous time is i and the hidden state at the next time is j1,o2，...,oT) The probability of (d);
Figure FDA0002660398870000049
indicating that, given HMM model parameters, the hidden state is i, and the observation sequence O ═ is present (O)1,o2，...,oT) The probability of (d);
3.2.3) B in B matrix elementsjkIs calculated iteratively
Figure FDA0002660398870000051
In the formula:
Figure FDA0002660398870000052
is the current estimated value of the HMM model parameter; the observation sequence O (O) is given in advance between 0 and T1,o2，...,oT) The hidden sequence I ═ I (I)1,i2,...iT) Wherein o isTIndicating the observed state at each time, iTRepresenting the implicit state result at each time instant;
Figure FDA0002660398870000053
indicating that, given HMM model parameters, the hidden state is i, and the observation sequence O ═ is present (O)1,o2，...,oT) The probability of (d); i (a)onK) is a decision factor, if and only if the observed state onK is 1, otherwise 0;
through the above calculation, the HMM model parameter λ of the current automobile can be obtained as (pi, a, B, n, m).
5. The intelligent vehicle lane change collision probability assessment method based on track prediction as claimed in claim 1, characterized in that: in step 5), calculating the relative positions of the target vehicle and the surrounding vehicles after t time according to the initial positions, the motion states and the prediction information of the vehicles, and determining collision critical positions so as to deduce the collision probability;
Figure FDA0002660398870000054
PH＝∑Pij
in the formula, DiIs the relative displacement between the target vehicle and the surrounding vehicle at time t, L0Denotes an initial longitudinal displacement difference between the target vehicle and the surrounding vehicle, H denotes a vehicle collision state, and when H is 0, i.e., Di>L0When the target vehicle collides with the surrounding vehicle; when H is 1, i.e. Di<L0When the target vehicle does not collide with the surrounding vehicle; pijThe predicted collision probability is calculated for each grid, and the relative displacement between the target vehicle and the surrounding vehicle is D after t timeiAnd Di>L0The probability of (c).
CN202010902992.2A 2020-09-01 2020-09-01 Intelligent vehicle lane change collision probability assessment method based on track prediction Active CN112258893B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010902992.2A CN112258893B (en) 2020-09-01 2020-09-01 Intelligent vehicle lane change collision probability assessment method based on track prediction

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010902992.2A CN112258893B (en) 2020-09-01 2020-09-01 Intelligent vehicle lane change collision probability assessment method based on track prediction

Publications (2)

Publication Number Publication Date
CN112258893A true CN112258893A (en) 2021-01-22
CN112258893B CN112258893B (en) 2021-10-26

Family

ID=74224575

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010902992.2A Active CN112258893B (en) 2020-09-01 2020-09-01 Intelligent vehicle lane change collision probability assessment method based on track prediction

Country Status (1)

Country Link
CN (1) CN112258893B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112965489A (en) * 2021-02-05 2021-06-15 北京理工大学 Intelligent vehicle high-speed lane change planning method based on collision detection
CN113916242A (en) * 2021-12-14 2022-01-11 腾讯科技（深圳）有限公司 Lane positioning method and device, storage medium and electronic equipment
CN114852099A (en) * 2021-02-03 2022-08-05 郑州宇通客车股份有限公司 Method for predicting lane changing behavior of motor vehicle
CN115331439A (en) * 2022-08-09 2022-11-11 山东旗帜信息有限公司 Vehicle history image-based highway interchange traffic flow prediction method and system

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106971194A (en) * 2017-02-16 2017-07-21 江苏大学 A kind of driving intention recognition methods based on the double-deck algorithms of improvement HMM and SVM
CN109969172A (en) * 2017-12-26 2019-07-05 华为技术有限公司 Control method for vehicle, equipment and computer storage medium
CN110085056A (en) * 2019-04-24 2019-08-02 华南理工大学 Vehicle lane-changing instantaneous risk recognition methods under a kind of highway bus or train route cooperative surroundings
CN110626339A (en) * 2019-07-10 2019-12-31 聊城大学 Emergency obstacle avoidance automatic driving assistance torque calculation method based on driver intention
CN110675656A (en) * 2019-09-24 2020-01-10 华南理工大学 Intelligent vehicle lane change early warning method based on instantaneous risk identification
CN110852281A (en) * 2019-11-13 2020-02-28 吉林大学 Driver lane change intention identification method based on Gaussian mixture hidden Markov model
CN111242484A (en) * 2020-01-14 2020-06-05 北京车汇天下科技有限公司 Vehicle risk comprehensive evaluation method based on transition probability
KR20200070854A (en) * 2018-12-10 2020-06-18 한국앱티브유한회사 Apparatus and method for identificating short cut-in vehicle and vehicle including the same
CN111301420A (en) * 2020-02-18 2020-06-19 北京汽车集团有限公司 Vehicle lane change control method and device, readable storage medium and vehicle
CN111445697A (en) * 2020-03-22 2020-07-24 华南理工大学 Expressway cooperative vehicle lane changing control method under intelligent network connection condition

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106971194A (en) * 2017-02-16 2017-07-21 江苏大学 A kind of driving intention recognition methods based on the double-deck algorithms of improvement HMM and SVM
CN109969172A (en) * 2017-12-26 2019-07-05 华为技术有限公司 Control method for vehicle, equipment and computer storage medium
KR20200070854A (en) * 2018-12-10 2020-06-18 한국앱티브유한회사 Apparatus and method for identificating short cut-in vehicle and vehicle including the same
CN110085056A (en) * 2019-04-24 2019-08-02 华南理工大学 Vehicle lane-changing instantaneous risk recognition methods under a kind of highway bus or train route cooperative surroundings
CN110626339A (en) * 2019-07-10 2019-12-31 聊城大学 Emergency obstacle avoidance automatic driving assistance torque calculation method based on driver intention
CN110675656A (en) * 2019-09-24 2020-01-10 华南理工大学 Intelligent vehicle lane change early warning method based on instantaneous risk identification
CN110852281A (en) * 2019-11-13 2020-02-28 吉林大学 Driver lane change intention identification method based on Gaussian mixture hidden Markov model
CN111242484A (en) * 2020-01-14 2020-06-05 北京车汇天下科技有限公司 Vehicle risk comprehensive evaluation method based on transition probability
CN111301420A (en) * 2020-02-18 2020-06-19 北京汽车集团有限公司 Vehicle lane change control method and device, readable storage medium and vehicle
CN111445697A (en) * 2020-03-22 2020-07-24 华南理工大学 Expressway cooperative vehicle lane changing control method under intelligent network connection condition

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
TING QU等: "Modeling driver"s car-following behavior based on hidden Markov model and model predictive control A cyber-physical system approach", 《2017 11TH ASIAN CONTROL CONFERENCE (ASCC)》 *
张可心: "智能汽车换道行为及危险性识别方法研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 *
李雅欣: "激光雷达建模与基于激光雷达的汽车行驶环境危险评估方法研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 *
杨诚: "基于隐马尔可夫理论的驾驶人换道意图识别研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 *
温惠英等: "高速公路入口匝道合流区的CP-CS融合模型", 《华南理工大学学报（自然科学版）》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114852099A (en) * 2021-02-03 2022-08-05 郑州宇通客车股份有限公司 Method for predicting lane changing behavior of motor vehicle
CN112965489A (en) * 2021-02-05 2021-06-15 北京理工大学 Intelligent vehicle high-speed lane change planning method based on collision detection
CN113916242A (en) * 2021-12-14 2022-01-11 腾讯科技（深圳）有限公司 Lane positioning method and device, storage medium and electronic equipment
CN113916242B (en) * 2021-12-14 2022-03-25 腾讯科技（深圳）有限公司 Lane positioning method and device, storage medium and electronic equipment
CN115331439A (en) * 2022-08-09 2022-11-11 山东旗帜信息有限公司 Vehicle history image-based highway interchange traffic flow prediction method and system
CN115331439B (en) * 2022-08-09 2023-08-18 山东旗帜信息有限公司 Expressway interchange traffic flow prediction method based on vehicle history image

Also Published As

Publication number Publication date
CN112258893B (en) 2021-10-26

Similar Documents

Publication Publication Date Title
CN112258893B (en) Intelligent vehicle lane change collision probability assessment method based on track prediction
CN110155046B (en) Automatic emergency braking hierarchical control method and system
CN110488802B (en) Decision-making method for dynamic behaviors of automatic driving vehicle in internet environment
EP3650297B1 (en) Method and apparatus for determining information related to a lane change of a target vehicle, and computer program
CN110843789B (en) Vehicle lane change intention prediction method based on time sequence convolution network
CN107848531B (en) Vehicle control device, vehicle control method, and medium storing vehicle control program
Murphey et al. Driver's style classification using jerk analysis
CN109808685B (en) Automobile early warning automatic collision avoidance control method based on danger assessment
CN109664882A (en) A kind of method avoiding road vehicle secondary collision, system and electronic equipment
CN112965489B (en) Intelligent vehicle high-speed lane change planning method based on collision detection
JPWO2017010344A1 (en) Vehicle control device, vehicle control method, and vehicle control program
CN110077398B (en) Risk handling method for intelligent driving
CN112249008B (en) Unmanned automobile early warning method aiming at complex dynamic environment
CN112285740A (en) Vehicle condition detection
CN113071482B (en) Intelligent vehicle anti-collision method based on passing time interval
Lengyel et al. Conflicts of automated driving with conventional traffic infrastructure
CN113962011A (en) Electric automobile braking system model and establishing method thereof
Oh et al. In-depth understanding of lane changing interactions for in-vehicle driving assistance systems
CN114132311B (en) Dangerous target screening method and module for automatic emergency braking of vehicle
CN114446046A (en) LSTM model-based weak traffic participant track prediction method
CN117068134A (en) Crossroad steering automatic braking system and method based on vehicle speed
CN114394091B (en) Vehicle speed control method for traffic vehicles in adaptive cruise system under parallel road scene
CN113830084B (en) Control method based on active collision avoidance of multi-lane vehicle and vehicle
CN114670851A (en) Driving assistance system, method, terminal and medium based on optimizing tracking algorithm
CN114872727A (en) Peripheral vehicle track prediction method applied to automatic driving vehicle

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