CN114379540B - Rollover-prevention driving decision method for large-sized operation vehicle considering influence of front obstacle - Google Patents
Rollover-prevention driving decision method for large-sized operation vehicle considering influence of front obstacle Download PDFInfo
- Publication number
- CN114379540B CN114379540B CN202210157766.5A CN202210157766A CN114379540B CN 114379540 B CN114379540 B CN 114379540B CN 202210157766 A CN202210157766 A CN 202210157766A CN 114379540 B CN114379540 B CN 114379540B
- Authority
- CN
- China
- Prior art keywords
- rollover
- vehicle
- driving
- decision
- function
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 91
- 230000002265 prevention Effects 0.000 claims abstract description 126
- 230000008569 process Effects 0.000 claims abstract description 28
- 230000001133 acceleration Effects 0.000 claims abstract description 17
- 238000012549 training Methods 0.000 claims abstract description 13
- 230000006870 function Effects 0.000 claims description 67
- 230000009471 action Effects 0.000 claims description 20
- 238000012546 transfer Methods 0.000 claims description 9
- 230000008901 benefit Effects 0.000 claims description 8
- 238000012512 characterization method Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 3
- 239000013598 vector Substances 0.000 claims description 3
- 238000009530 blood pressure measurement Methods 0.000 claims 1
- 230000007547 defect Effects 0.000 abstract description 2
- 206010039203 Road traffic accident Diseases 0.000 description 6
- 230000000694 effects Effects 0.000 description 6
- 230000002787 reinforcement Effects 0.000 description 6
- 238000011160 research Methods 0.000 description 4
- 238000011217 control strategy Methods 0.000 description 2
- 238000003912 environmental pollution Methods 0.000 description 2
- 238000004880 explosion Methods 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000011158 quantitative evaluation Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
Classifications
-
- 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, or advanced driver assistance systems for ensuring comfort, stability and safety or drive control systems for propelling or retarding the vehicle
- B60W30/02—Control of vehicle driving stability
- B60W30/04—Control of vehicle driving stability related to roll-over prevention
-
- 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, or advanced driver assistance systems for ensuring comfort, stability and safety or drive control systems for propelling or retarding the vehicle
- B60W30/02—Control of vehicle driving stability
- B60W30/04—Control of vehicle driving stability related to roll-over prevention
- B60W2030/043—Control of vehicle driving stability related to roll-over prevention about the roll axis
-
- 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
- B60W2552/00—Input parameters relating to infrastructure
- B60W2552/50—Barriers
Abstract
The invention discloses a rollover-prevention driving decision method for a large-sized operating vehicle, which considers the influence of a front obstacle. Firstly, a traffic scene suitable for the rollover prevention driving decision method is defined. And modeling the rollover prevention driving decision problem as a Markov decision process, and establishing a rollover prevention driving decision model of the large-scale operating vehicle by using a standard strategy gradient algorithm. And finally, training the rollover prevention driving decision model to obtain rollover prevention driving strategies under different driving conditions. The decision method provided by the invention considers potential safety hazards in the process of executing the rollover prevention driving strategy, overcomes the defect that the existing method lacks of effectiveness and reliability, provides clear driving strategies such as straight running, steering, acceleration, deceleration and the like for the commercial vehicle, and realizes effective and reliable rollover prevention driving decision of the large-scale commercial vehicle.
Description
Technical Field
The invention relates to a rollover prevention driving decision method, in particular to a rollover prevention driving decision method for a large-sized operating vehicle considering the influence of a front obstacle, and belongs to the technical field of automobile safety.
Background
The operating vehicle is used as a main undertaker of road transportation in China, and the safety condition of the operating vehicle directly influences the road transportation safety in China. In case of traffic accidents in the transportation process, serious consequences such as group death, falling of goods, burning, explosion and the like are easily caused, bad effects such as property loss, environmental pollution, ecological damage and the like are caused, and large-scale and extra-large-scale safety accidents are easily induced, so that road traffic safety is seriously threatened.
Related data of the U.S. highway traffic safety authorities indicate that the degree of damage to rollover accidents is next to collision accidents in all commercial vehicle traffic accidents, with the 2 nd position. Therefore, the active rollover prevention and control has important significance for guaranteeing road traffic safety and improving the prevention and control capability of serious accidents of road transportation. The rollover prevention driving decision is taken as an important ring of rollover active prevention and control, if a reasonable rollover prevention intervention strategy can be determined before a rollover accident occurs, accurate and reliable braking and deceleration driving strategies are provided for operating vehicles, and the occurrence frequency of traffic accidents caused by rollover can be greatly reduced.
At present, literature and patent research on a rollover prevention driving decision method mainly comprises a dynamic model-based method and a data driving-based method. The rollover prevention decision method based on the dynamic model is to identify the vehicle rollover risk by using parameters such as lateral acceleration, roll angle and the like, take the wheel braking force or steering angle as control quantity, and establish a rollover prevention control strategy by using control algorithms such as PID (Proportional INTEGRAL DERIVATIVE) control, sliding mode control and the like. However, the dynamics models of different vehicles have variability, making rollover prevention decisions less reliable and adaptable.
The rollover prevention decision method based on data driving is to extract and analyze data such as vehicle motion state information, train and fit a model through an interactive self-learning mechanism, and further form an automatic decision model, and is rarely used at present. The invention relates to a heavy-duty commercial vehicle rollover-prevention driving decision method based on deep reinforcement learning (publication number: CN 112580148A) and a large-scale commercial vehicle rollover-prevention decision method considering road surface attachment conditions (application number: 202111225841.9), which are used for establishing a rollover-prevention driving decision model by utilizing a deep reinforcement learning algorithm to obtain a rollover-prevention driving strategy of the commercial vehicle. However, the traffic environment to which the above patent is applicable is relatively simple, and the influence of the preceding vehicle on the rollover prevention driving decision is not considered.
In general, although the above two methods have a certain rollover prevention effect, there is still a disadvantage in terms of effectiveness and reliability of decision making, because potential safety hazards in executing the rollover prevention driving strategy are not considered. Particularly, in a traffic environment where an obstacle exists in front of a commercial vehicle, if rollover prevention driving strategies such as release braking are adopted, a longer braking distance is likely to increase the risk of forward collision. If emergency braking is performed, forward collision can be avoided, but excessive braking deceleration further aggravates the rollover risk of the vehicle, so that the operating vehicle is more easily unstable to rollover, and the rollover prevention effect cannot be achieved. Therefore, reasonable and effective rollover prevention driving decision should be taken into consideration both the forward collision prevention and rollover prevention.
The existing rollover-preventing driving decision method for the large-scale commercial vehicle does not consider the influence of a front obstacle on driving safety, ignores potential safety hazards in the process of executing a rollover-preventing driving strategy, and cannot meet the driving safety requirements of simultaneously avoiding forward collision and vehicle rollover. Under the traffic environment that the obstacle exists in front of the commercial vehicle, an effective and reliable rollover driving prevention decision method for the large-scale commercial vehicle is not available.
Disclosure of Invention
The invention aims to: aiming at the problem that the rollover prevention driving decision method of the large-scale commercial vehicle is low in effectiveness and reliability, the invention discloses the rollover prevention driving decision method of the large-scale commercial vehicle based on consideration of the influence of a front obstacle. The method considers the potential safety hazards in the process of executing the rollover prevention driving strategy, overcomes the defect that the prior method lacks of effectiveness and reliability, provides clear driving strategies such as straight running, steering, acceleration, deceleration and the like for the commercial vehicles, effectively reduces the potential safety hazards in the process of executing the rollover prevention driving strategy, and improves the effectiveness and reliability of rollover prevention driving decisions of the large commercial vehicles
The technical scheme is as follows: the invention provides an anti-rollover driving decision method considering the influence of a front obstacle for a large-scale operation vehicle, such as a semi-trailer tank car and a heavy truck. Firstly, a traffic scene suitable for the rollover prevention driving decision method is defined. And modeling the rollover prevention driving decision problem as a Markov decision process, and establishing a rollover prevention driving decision model of the large-scale operating vehicle by using a standard strategy gradient algorithm. Finally, training the rollover-preventing driving decision model to obtain rollover-preventing driving strategies under different driving conditions, so that effective and reliable rollover-preventing driving decision is realized. The method comprises the following steps:
Step one: traffic scene suitable for clear rollover prevention driving decision method
In order to reduce or avoid rollover accidents of large-scale operation vehicles and improve the operation safety of the vehicles, the invention provides a rollover active safety prevention and control decision method of the large-scale operation vehicles, which considers the influence of front obstacles, and the method is applicable to the following scenes:
The large commercial vehicle (i.e., host vehicle C 0) is traveling on a multi-lane highway with traffic participants (i.e., lead vehicle C 1) in front of the lane. When the self-vehicle brakes, changes lanes or passes through a curve, decision strategies such as braking deceleration, steering and the like are provided for a driver effectively and timely in order to ensure driving safety.
In the present invention, multilane means that the number of lanes is 3 or more. The front vehicle is a vehicle which is located in front of a road on which a large-sized operating vehicle (own vehicle) travels, and which is located in the same lane line, has the same traveling direction, and is closest to the road.
Step two: establishing rollover-preventing driving decision model of large-scale operation vehicle
Aiming at the problem of low effectiveness and reliability of the rollover-prevention driving decision method of the large-scale commercial vehicle, the invention comprehensively considers the influence of the operation of a driver, the running working condition and the traffic environment on the driving decision, and establishes an effective and reliable rollover-prevention driving decision model.
The invention adopts a standard strategy gradient algorithm, establishes a rollover prevention driving decision model based on the traffic scene in the step one, and researches the rollover prevention driving strategy of the operating vehicle under the condition that the front obstacle exists. The method specifically comprises the following 4 substeps:
Sub-step 1: basic parameters defining rollover prevention driving decision model
Considering that the future motion state of a large-sized operating vehicle is influenced by the current motion state and the current motion at the same time, the invention models the rollover prevention driving decision problem as a Markov decision process and defines basic parameters of the model: state space S t at time t, state space S t+1 at time t+1, action space a t at time t, and return value R t corresponding to action space a t. Specifically:
(1) Defining a state space
Roll stability of large commercial vehicles is related not only to the state of motion of the vehicle itself but also to the road state. Thus, the present invention defines a state space using vehicle motion state information:
St＝[vlon,vhor,alon,ahor,ωyaw,θroll,θswa,δbrake,δthr,Lf,Drel] (1)
Wherein v lon,vhor represents the longitudinal speed and the transverse speed of the large-scale operation vehicle respectively, and the units are meters per second; a lon,ahor represents longitudinal acceleration and lateral acceleration respectively, and the unit is square seconds every second meter, and can be obtained through measurement of a centimeter-level high-precision integrated navigation system; omega yaw,θroll respectively represents yaw rate and roll angle, and the units are radian per second and degree respectively, and can be obtained through measurement of an MEMS gyroscope; theta swa is the steering wheel angle of the vehicle, delta brake,δthr is the opening degree of a brake pedal and the opening degree of a throttle valve respectively, and the units are percentages, and CAN be obtained by reading the CAN bus information of the vehicle body; l f is the transverse transfer rate of the leaf spring pressure, and can be obtained through calculation of the pressure borne by the axle leaf spring measured by the pressure sensor, D rel is the relative distance between the vehicle and the front vehicle, the unit is meter, and the axle leaf spring can be obtained through millimeter wave radar acquisition.
(2) Defining an action space
In order to comprehensively consider the influence of transverse and longitudinal control on rollover prevention and reasonably and effectively output a rollover prevention driving decision strategy, the invention takes steering wheel rotation angle and brake pedal opening as control amounts, and defines the driving strategy output by a decision model, namely an action space A t＝[θswa_out,δbrake_out at the moment t.
Wherein, θ swa_out represents the normalized steering wheel angle control amount, the range is [ -1,1], and δ brake_out represents the normalized brake pedal opening, the range is [0,1]. When δ brake =0, it means that the vehicle is not braked, and when δ brake =1, it means that the vehicle is braked at the maximum braking deceleration.
(3) Defining a reward function
In order to realize quantitative evaluation of the advantages and disadvantages of the action space, the invention designs the reward function as follows:
Rt＝δ1·r1(t)+δ2·r2(t)+r3(t) (2)
Wherein R t is a total rewarding function at the moment t, and R 1(t),r2(t),r3 (t) respectively represents a side-turning preventing rewarding function, a rear-end collision preventing rewarding function and a punishment function; delta 1,δ2 represents the weight coefficient of the rollover prevention bonus function and the weight coefficient of the rear-end collision prevention bonus function, respectively.
Considering the problem of lack of accuracy in estimating rollover risk by utilizing a single rollover characterization parameter, comprehensively considering the influence of 3 characterization parameters of the roll angle, the lateral acceleration and the leaf spring pressure lateral load transfer rate on rollover, and establishing an anti-rollover reward function r 1 (t):
Wherein a thr,θthr,Lthr represents a preset lateral acceleration threshold value, a preset roll angle threshold value and a preset leaf spring pressure lateral transfer rate threshold value, and mu 1 represents a weight coefficient of a reward function r 1 (t);
In order to reduce potential safety hazards in the process of executing the rollover prevention driving strategy, a large-sized operating vehicle should be prevented from colliding with a front obstacle in the rollover prevention decision process; to this end, a forward anti-collision reward function r 2 (t) is established:
r2(t)＝μ2|Drel-Dsafe| (4)
Wherein D safe represents the safe distance between the own vehicle and the preceding vehicle, and μ 2 represents the weight coefficient of the reward function r 2 (t);
Considering reasonable safety distance, the traffic efficiency and the driving safety are considered at the same time; for this purpose, a variable headway is used as the minimum safety distance D w for the autonomous operating vehicle;
Dw＝vlonτ+vforT+Lmin (5)
Where τ represents the headway in seconds, v for represents the speed of the vehicle ahead in meters per second, and L min represents the critical distance in meters;
to correct the error strategy in the driving decision process, a penalty function is established:
r3(t)＝-Spen (6)
In the formula, S pen is a penalty value, and in the invention, S pen =200 is taken to represent that if a rollover or rear-end collision accident occurs to the vehicle, the decision model obtains a penalty of-200.
Sub-step 2: network architecture for designing rollover prevention driving decision model
And constructing a rollover prevention driving decision model by using an Actor-Critic framework, wherein the decision model comprises an Actor network and a Critic network. The Actor network takes a state space S t as input and carries out regression on the feature vectors, so that continuous action A t is output; the Critic network takes as input the state space S t and action a t to evaluate the value of the current "state-action".
Three layers of full-connection networks with the same structure are built for the Actor and the Critic networks, the activation functions of the three layers of networks are all linear rectification functions (RECTIFIED LINEAR Unit, reLU), and the expression is as follows: f (x) =max (0, x).
Step three: training rollover prevention driving decision model
Training parameters in the rollover prevention driving decision model, wherein the method specifically comprises the following steps of:
Sub-step 1: initializing a parameter theta 0 of the strategy function and a parameter phi 0 of the value function;
Sub-step 2: iteratively updating parameters of the policy function and the value function, each iteration comprising sub-steps 2.1 to 2.6, in particular:
Substep 2.1: executing a policy pi k＝π(θk in the environment), gathering a set of tracks D k＝{τi;
substep 2.2: calculating a subsequent discount prize value
Substep 2.3: based on the current value functionTaking the time sequence difference algorithm as an estimated value of the advantage function, and calculating an estimated value of the advantage function;
Substep 2.4: estimating a strategy gradient;
substep 2.5: updating a calculation strategy;
Substep 2.6: fitting a value function;
Sub-step 3: and (3) carrying out iterative updating according to the methods provided in the substep 1 and the substep 2, so that the rollover prevention driving decision model gradually converges. In the training process, if the vehicle turns over or collides, the current round is terminated and a new round is started to train. And when the iteration reaches the maximum times or the large-scale operation vehicle stably and effectively realizes rollover prevention by utilizing the decision strategy output by the model, indicating that the iteration is completed.
Finally, the motion state information of the large-scale operation vehicle is input into the trained rollover-preventing driving decision model, so that the rollover-preventing driving decision strategy can be output on line, and the effective and reliable rollover-preventing driving decision of the large-scale operation vehicle is realized.
The beneficial effects are that: compared with a general rollover prevention driving decision method, the method provided by the invention has the characteristics of more effectiveness and reliability, and is specifically embodied as follows:
(1) According to the method provided by the invention, a refined reward function is designed aiming at the dynamic coupling of the rollover prevention elements in the driving decision process, the influence of factors such as a driving safety threshold value, a forward obstacle type, vehicle rolling stability and the like on the driving safety of the commercial vehicle is comprehensively considered, and rollover prevention driving strategies such as straight running, steering, acceleration, deceleration and the like under different driving working conditions can be output, so that the effective and reliable rollover prevention driving decision of the commercial vehicle is realized.
(2) The method provided by the invention considers the influence of the front obstacle on the driving safety, and simultaneously considers two aspects of rollover prevention and forward collision, so that the potential safety hazard in the process of executing the rollover prevention driving strategy can be effectively reduced, and the effectiveness and reliability of rollover prevention driving decisions of the large-scale operating vehicle are further improved.
(3) The method provided by the invention does not need to carry out complex dynamic modeling, and the calculation method is simple and clear.
Drawings
FIG. 1 is a schematic illustration of a technical route of the present invention;
Fig. 2 is a schematic view of a traffic scenario in which the present invention is applicable.
Detailed Description
The technical scheme of the invention is further described below with reference to the accompanying drawings.
The operating vehicle is used as a main undertaker of road transportation in China, and the safety condition of the operating vehicle directly influences the road transportation safety in China. In case of traffic accidents in the transportation process, serious consequences such as group death, falling of goods, burning, explosion and the like are easily caused, bad effects such as property loss, environmental pollution, ecological damage and the like are caused, and large-scale and extra-large-scale safety accidents are easily induced, so that road traffic safety is seriously threatened.
Related data of the U.S. highway traffic safety authorities indicate that the degree of damage to rollover accidents is next to collision accidents in all commercial vehicle traffic accidents, with the 2 nd position. Therefore, the active rollover prevention and control has important significance for guaranteeing road traffic safety and improving the prevention and control capability of serious accidents of road transportation. The rollover prevention driving decision is taken as an important ring of rollover active prevention and control, if a reasonable rollover prevention intervention strategy can be determined before a rollover accident occurs, accurate and reliable braking and deceleration driving strategies are provided for operating vehicles, and the occurrence frequency of traffic accidents caused by rollover can be greatly reduced.
At present, literature and patent research on a rollover prevention driving decision method mainly comprises a dynamic model-based method and a data driving-based method. The rollover prevention decision method based on the dynamic model is to identify the vehicle rollover risk by using parameters such as lateral acceleration, roll angle and the like, take the wheel braking force or steering angle as control quantity, and establish a rollover prevention control strategy by using control algorithms such as PID (Proportional INTEGRAL DERIVATIVE) control, sliding mode control and the like. However, the dynamics models of different vehicles have variability, making rollover prevention decisions less reliable and adaptable.
The rollover prevention decision method based on data driving is to extract and analyze data such as vehicle motion state information, train and fit a model through an interactive self-learning mechanism, and further form an automatic decision model, and is rarely used at present. The invention relates to a heavy-duty commercial vehicle rollover-prevention driving decision method based on deep reinforcement learning (publication number: CN 112580148A) and a large-scale commercial vehicle rollover-prevention decision method considering road surface attachment conditions (application number: 202111225841.9), which are used for establishing a rollover-prevention driving decision model by utilizing a deep reinforcement learning algorithm to obtain a rollover-prevention driving strategy of the commercial vehicle. However, the traffic environment to which the above patent is applicable is relatively simple, and the influence of the preceding vehicle on the rollover prevention driving decision is not considered.
In general, although the above two methods have a certain rollover prevention effect, there is still a disadvantage in terms of effectiveness and reliability of decision making, because potential safety hazards in executing the rollover prevention driving strategy are not considered. Particularly, in a traffic environment where an obstacle exists in front of a commercial vehicle, if rollover prevention driving strategies such as release braking are adopted, a longer braking distance is likely to increase the risk of forward collision. If emergency braking is performed, forward collision can be avoided, but excessive braking deceleration further aggravates the rollover risk of the vehicle, so that the operating vehicle is more easily unstable to rollover, and the rollover prevention effect cannot be achieved. Therefore, reasonable and effective rollover prevention driving decision should be taken into consideration both the forward collision prevention and rollover prevention.
The existing rollover-preventing driving decision method for the large-scale commercial vehicle does not consider the influence of a front obstacle on driving safety, ignores potential safety hazards in the process of executing a rollover-preventing driving strategy, and cannot meet the driving safety requirements of simultaneously avoiding forward collision and vehicle rollover. Under the traffic environment that the obstacle exists in front of the commercial vehicle, an effective and reliable rollover driving prevention decision method for the large-scale commercial vehicle is not available.
Aiming at the problem of low effectiveness and reliability of a rollover prevention driving decision method of a large-sized operation vehicle, the invention provides the rollover prevention driving decision method of the large-sized operation vehicle considering the influence of a front obstacle aiming at the large-sized operation vehicle such as a semi-trailer tank car and a heavy truck. Firstly, a traffic scene suitable for the rollover prevention driving decision method is defined. And modeling the rollover prevention driving decision problem as a Markov decision process, and establishing a rollover prevention driving decision model of the large-scale operating vehicle by using a standard strategy gradient algorithm. Finally, training the rollover prevention driving decision model to obtain rollover prevention driving strategies under different driving conditions, so that the rollover prevention driving decision of the large-scale operating vehicle is effectively and reliably realized. The technical route of the invention is shown in figure 1, and the specific steps are as follows:
Step one: traffic scene suitable for clear rollover prevention driving decision method
In order to reduce or avoid rollover accidents of large-scale operation vehicles and improve operation safety of the vehicles, the invention provides a rollover active safety prevention and control decision method of the large-scale operation vehicles, which considers the influence of front obstacles, and the applicable scene is shown in fig. 2, and is specifically described as follows:
The large commercial vehicle (i.e., host vehicle C 0) is traveling on a multi-lane highway with traffic participants (i.e., lead vehicle C 1) in front of the lane. When the self-vehicle brakes, changes lanes or passes through a curve, decision strategies such as braking deceleration, steering and the like are provided for a driver effectively and timely in order to ensure driving safety.
In the present invention, multilane means that the number of lanes is 3 or more. The front vehicle is a vehicle which is located in front of a road on which a large-sized operating vehicle (own vehicle) travels, and which is located in the same lane line, has the same traveling direction, and is closest to the road.
Step two: establishing rollover-preventing driving decision model of large-scale operation vehicle
Aiming at the problem of low effectiveness and reliability of the rollover-prevention driving decision method of the large-scale commercial vehicle, the invention comprehensively considers the influence of the operation of a driver, the running working condition and the traffic environment on the driving decision, and establishes an effective and reliable rollover-prevention driving decision model.
The complexity and uncertainty of the road state and the driver behavior are important factors influencing the rollover prevention driving decision, and the rollover prevention driving decision model is established by adopting a deep reinforcement learning algorithm in consideration of the adaptability characteristics of deep reinforcement learning to the uncertainty and the full mining and characterization capability of high-dimensional characteristics such as the road state and the like.
The invention adopts a standard strategy gradient algorithm, establishes a rollover prevention driving decision model based on the traffic scene in the step one, and researches the rollover prevention driving strategy of the operating vehicle under the condition that the front obstacle exists. The method specifically comprises the following 4 substeps:
Sub-step 1: basic parameters defining rollover prevention driving decision model
Considering that the future motion state of a large-sized operating vehicle is influenced by the current motion state and the current motion at the same time, the invention models the rollover prevention driving decision problem as a Markov decision process and defines basic parameters of the model: state space S t at time t, state space S t+1 at time t+1, action space a t at time t, and return value R t corresponding to action space a t. Specifically:
(1) Defining a state space
Roll stability of large commercial vehicles is related not only to the state of motion of the vehicle itself but also to the road state. Thus, the present invention defines a state space using vehicle motion state information:
St＝[vlon,vhor,alon,ahor,ωyaw,θroll,θswa,δbrake,δthr,Lf,Drel] (1)
Wherein v lon,vhor represents the longitudinal speed and the transverse speed of the large-scale operation vehicle respectively, and the units are meters per second; a lon,ahor represents longitudinal acceleration and lateral acceleration respectively, and the unit is square seconds every second meter, and can be obtained through measurement of a centimeter-level high-precision integrated navigation system; omega yaw,θroll respectively represents yaw rate and roll angle, and the units are radian per second and degree respectively, and can be obtained through measurement of an MEMS gyroscope; theta swa is the steering wheel angle of the vehicle, delta brake,δthr is the opening degree of a brake pedal and the opening degree of a throttle valve respectively, and the units are percentages, and CAN be obtained by reading the CAN bus information of the vehicle body; l f is the transverse transfer rate of the leaf spring pressure, and can be obtained through calculation of the pressure borne by the axle leaf spring measured by the pressure sensor, D rel is the relative distance between the vehicle and the front vehicle, the unit is meter, and the axle leaf spring can be obtained through millimeter wave radar acquisition.
(2) Defining an action space
In order to comprehensively consider the influence of transverse and longitudinal control on rollover prevention and reasonably and effectively output a rollover prevention driving decision strategy, the invention takes steering wheel rotation angle and brake pedal opening as control amounts, and defines the driving strategy output by a decision model, namely an action space A t＝[θswa_out,δbrake_out at the moment t.
Wherein, θ swa_out represents the normalized steering wheel angle control amount, the range is [ -1,1], and δ brake_out represents the normalized brake pedal opening, the range is [0,1]. When δ brake =0, it means that the vehicle is not braked, and when δ brake =1, it means that the vehicle is braked at the maximum braking deceleration.
(3) Defining a reward function
In order to realize quantitative evaluation of the advantages and disadvantages of the action space, the invention designs the reward function as follows:
Rt＝δ1·r1(t)+δ2·r2(t)+r3(t) (2)
Wherein R t is a total rewarding function at the moment t, and R 1(t),r2(t),r3 (t) respectively represents a side-turning preventing rewarding function, a rear-end collision preventing rewarding function and a punishment function; delta 1,δ2 represents the weight coefficient of the rollover prevention bonus function and the weight coefficient of the rear-end collision prevention bonus function, respectively.
Considering the problem of lack of accuracy in estimating rollover risk by utilizing a single rollover characterization parameter, comprehensively considering the influence of 3 characterization parameters of the roll angle, the lateral acceleration and the leaf spring pressure lateral load transfer rate on rollover, and establishing an anti-rollover reward function r 1 (t):
Wherein a thr,θthr,Lthr represents a preset lateral acceleration threshold value, a preset roll angle threshold value and a preset leaf spring pressure lateral transfer rate threshold value, and mu 1 represents a weight coefficient of a reward function r 1 (t);
In order to reduce potential safety hazards in the process of executing the rollover prevention driving strategy, a large-sized operating vehicle should be prevented from colliding with a front obstacle in the rollover prevention decision process; to this end, a forward anti-collision reward function r 2 (t) is established:
r2(t)＝μ2|Drel-Dsafe| (4)
Wherein D safe represents the safe distance between the own vehicle and the preceding vehicle, and μ 2 represents the weight coefficient of the reward function r 2 (t);
Considering reasonable safety distance, the traffic efficiency and the driving safety are considered at the same time; for this purpose, a variable headway is used as the minimum safety distance D w for the autonomous operating vehicle;
Dw＝vlonτ+vforT+Lmin (5)
Where τ represents the headway in seconds, v for represents the speed of the vehicle ahead in meters per second, and L min represents the critical distance in meters;
to correct the error strategy in the driving decision process, a penalty function is established:
r3(t)＝-Spen (6)
In the formula, S pen is a penalty value, and in the invention, S pen =200 is taken to represent that if a rollover or rear-end collision accident occurs to the vehicle, the decision model obtains a penalty of-200.
Sub-step 2: network architecture for designing rollover prevention driving decision model
And constructing a rollover prevention driving decision model by using an Actor-Critic framework, wherein the decision model comprises an Actor network and a Critic network. The Actor network takes a state space S t as input and carries out regression on the feature vectors, so that continuous action A t is output; the Critic network takes as input the state space S t and action a t to evaluate the value of the current "state-action".
Three layers of full-connection networks with the same structure are built for the Actor and the Critic networks, the activation functions of the three layers of networks are all linear rectification functions (RECTIFIED LINEAR Unit, reLU), and the expression is as follows: f (x) =max (0, x).
Step three: training rollover prevention driving decision model
Training parameters in the rollover prevention driving decision model, wherein the method specifically comprises the following steps of:
Sub-step 1: initializing a parameter theta 0 of the strategy function and a parameter phi 0 of the value function;
Sub-step 2: iteratively updating parameters of the policy function and the value function, each iteration comprising sub-steps 2.1 to 2.6, in particular:
Substep 2.1: executing a policy pi k＝π(θk in the environment), gathering a set of tracks D k＝{τi;
substep 2.2: calculating a subsequent discount prize value
Substep 2.3: based on the current value functionTaking the time sequence difference algorithm as an estimated value of the advantage function, and calculating an estimated value of the advantage function;
Substep 2.4: estimating a strategy gradient;
substep 2.5: updating a calculation strategy;
Substep 2.6: fitting a value function;
Sub-step 3: and (3) carrying out iterative updating according to the methods provided in the substep 1 and the substep 2, so that the rollover prevention driving decision model gradually converges. In the training process, if the vehicle turns over or collides, the current round is terminated and a new round is started to train. And when the iteration reaches the maximum times or the large-scale operation vehicle stably and effectively realizes rollover prevention by utilizing the decision strategy output by the model, indicating that the iteration is completed.
Finally, the motion state information of the large-scale operation vehicle is input into the trained rollover-preventing driving decision model, so that the rollover-preventing driving decision strategy can be output on line, and the effective and reliable rollover-preventing driving decision of the large-scale operation vehicle is realized.
Claims (1)
1. A large-scale operation vehicle rollover prevention driving decision method considering the influence of a front obstacle; firstly, defining a traffic scene to which the rollover active prevention and control decision method is applicable; secondly, acquiring motion state information of the vehicle by using a sensor; finally, modeling the rollover active prevention and control decision problem as a Markov decision process, and establishing a rollover active prevention and control decision model of the large-scale operating vehicle by using a standard strategy gradient algorithm to obtain rollover prevention driving strategies under different driving conditions; the method is characterized in that:
Step one: traffic scene suitable for clear rollover prevention driving decision method
The large operating vehicle, namely the self vehicle C 0 runs on the multi-lane high-grade highway, and a traffic participant, namely the front vehicle C 1 exists in front of the lane; when the self-vehicle brakes, changes lanes or passes through a curve, a decision strategy comprising braking, decelerating and steering is provided for a driver effectively and timely in order to ensure driving safety;
Multilane refers to the number of lanes being 3 or more; the front vehicle is a vehicle which is positioned in front of a running road of the own vehicle C 0, positioned in the same lane line, has the same running direction and has the nearest distance;
Step two: establishing rollover-preventing driving decision model of large-scale operation vehicle
Adopting a standard strategy gradient algorithm, establishing an anti-rollover driving decision model based on the traffic scene in the step one, and researching an anti-rollover driving strategy of the operating vehicle under the condition that a front obstacle exists; the method specifically comprises the following 4 substeps:
Sub-step 1: basic parameters defining rollover prevention driving decision model
Considering that the future motion state of a large-sized operating vehicle is influenced by the current motion state and the current motion at the same time, modeling the rollover prevention driving decision problem as a Markov decision process, and defining basic parameters of the model: state space S t at time t, state space S t+1 at time t+1, action space a t at time t, and return value R t corresponding to action space a t; specifically:
(1) Defining a state space
Roll stability of large commercial vehicles is related not only to the motion state of the vehicle itself but also to the road state; thus, a state space is defined using the vehicle motion state information:
St＝[vlon,vhor,alon,ahor,ωyaw,θroll,θswa,δbrake,δthr,Lf,Drel] (1)
Wherein v lon,vhor represents the longitudinal speed and the transverse speed of the large-scale operation vehicle respectively, and the units are meters per second; a lon,ahor respectively represents longitudinal acceleration and lateral acceleration, the units are square seconds every meter, and the unit is obtained through measurement of a centimeter-level high-precision integrated navigation system; omega yaw,θroll respectively represents yaw rate and roll angle, and units are radian per second and degree respectively, and are obtained through measurement of an MEMS gyroscope; theta swa is the steering wheel angle of the vehicle, delta brake,δthr is the opening degree of a brake pedal and the opening degree of a throttle valve respectively, and the units are percentages, and the vehicle body CAN bus information is read to obtain the vehicle body CAN bus information; l f is the transverse transfer rate of the leaf spring pressure, the leaf spring pressure of the axle is calculated and obtained through pressure measurement of a pressure sensor, D rel is the relative distance between the vehicle and the front vehicle, the unit is meter, and the leaf spring pressure is obtained through millimeter wave radar acquisition;
(2) Defining an action space
The steering wheel angle and the opening degree of a brake pedal are used as control amounts, and a driving strategy output by a decision model, namely an action space A t＝[θswa_out,δbrake_out at the moment t, is defined;
wherein, θ swa_out represents the normalized steering wheel angle control amount, the range is [ -1,1], and δ brake_out represents the normalized brake pedal opening, the range is [0,1]; when δ brake =0, it means that the vehicle is not braked, and when δ brake =1, it means that the vehicle is braked at the maximum braking deceleration;
(3) Defining a reward function
The bonus function is designed to:
Rt＝δ1·r1(t)+δ2·r2(t)+r3(t) (2)
Wherein R t is a total rewarding function at the moment t, and R 1(t),r2(t),r3 (t) respectively represents a side-turning preventing rewarding function, a rear-end collision preventing rewarding function and a punishment function; delta 1,δ2 represents the weight coefficient of the rollover prevention reward function and the weight coefficient of the rear-end collision prevention reward function respectively;
Considering the problem of lack of accuracy in estimating rollover risk by utilizing a single rollover characterization parameter, comprehensively considering the influence of 3 characterization parameters of the roll angle, the lateral acceleration and the leaf spring pressure lateral load transfer rate on rollover, and establishing an anti-rollover reward function r 1 (t):
Wherein a thr,θthr,Lthr represents a preset lateral acceleration threshold value, a preset roll angle threshold value and a preset leaf spring pressure lateral transfer rate threshold value, and mu 1 represents a weight coefficient of a reward function r 1 (t);
In order to reduce potential safety hazards in the process of executing the rollover prevention driving strategy, a large-sized operating vehicle should be prevented from colliding with a front obstacle in the rollover prevention decision process; to this end, a forward anti-collision reward function r 2 (t) is established:
r2(t)＝μ2|Drel-Dsafe| (4)
Wherein D safe represents the safe distance between the own vehicle and the preceding vehicle, and μ 2 represents the weight coefficient of the reward function r 2 (t);
Considering reasonable safety distance, the traffic efficiency and the driving safety are considered at the same time; for this purpose, a variable headway is used as the minimum safety distance D w for the autonomous operating vehicle;
Dw＝vlonτ+vforT+Lmin (5)
Where τ represents the headway in seconds, v for represents the speed of the vehicle ahead in meters per second, and L min represents the critical distance in meters;
To correct the error strategy in the driving decision process, a penalty function r 3 (t) is established:
r3(t)＝-Spen (6)
Wherein, S pen is a punishment value, S pen =200 is taken to represent that if the vehicle has rollover or rear-end collision accidents, the decision model obtains punishment of-200;
Sub-step 2: network architecture for designing rollover prevention driving decision model
Constructing a rollover prevention driving decision model by using an Actor-Critic framework, wherein the decision model comprises an Actor network and a Critic network; the Actor network takes a state space S t as input and carries out regression on the feature vectors, so that continuous action A t is output; the Critic network takes as input the state space S t and action a t, evaluating the value of the current "state-action";
three layers of full-connection networks with the same structure are built for the Actor and the Critic networks, the activation functions of the three layers of networks are all linear rectification functions, and the expression is as follows: f (x) =max (0, x);
Step three: training rollover prevention driving decision model
Training parameters in the rollover prevention driving decision model, wherein the method specifically comprises the following steps of:
Sub-step 1: initializing a parameter theta 0 of the strategy function and a parameter phi 0 of the value function;
Sub-step 2: iteratively updating parameters of the policy function and the value function, each iteration comprising sub-steps 2.1 to 2.6, in particular:
Substep 2.1: executing a policy pi k＝π(θk in the environment), gathering a set of tracks D k＝{τi;
substep 2.2: calculating a subsequent discount prize value
Substep 2.3: based on the current value functionTaking the time sequence difference algorithm as an estimated value of the advantage function, and calculating an estimated value of the advantage function;
Substep 2.4: estimating a strategy gradient;
substep 2.5: updating a calculation strategy;
Substep 2.6: fitting a value function;
Sub-step 3: iterative updating is carried out according to the methods provided by the substep 1 and the substep 2, so that the rollover prevention driving decision model gradually converges; in the training process, if the vehicle turns over or collides, stopping the current round and starting a new round for training; when iteration reaches the maximum times or the large-scale operation vehicle stably and effectively realizes rollover prevention by utilizing a decision strategy output by the model, indicating that the iteration is completed;
finally, the motion state information of the large-scale operation vehicle is input into the trained rollover-preventing driving decision model, so that the rollover-preventing driving decision strategy can be output on line, and the effective and reliable rollover-preventing driving decision of the large-scale operation vehicle is realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210157766.5A CN114379540B (en) | 2022-02-21 | 2022-02-21 | Rollover-prevention driving decision method for large-sized operation vehicle considering influence of front obstacle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210157766.5A CN114379540B (en) | 2022-02-21 | 2022-02-21 | Rollover-prevention driving decision method for large-sized operation vehicle considering influence of front obstacle |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114379540A CN114379540A (en) | 2022-04-22 |
| CN114379540B true CN114379540B (en) | 2024-04-30 |
Family
ID=81206335
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210157766.5A Active CN114379540B (en) | 2022-02-21 | 2022-02-21 | Rollover-prevention driving decision method for large-sized operation vehicle considering influence of front obstacle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114379540B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114863708B (en) * | 2022-05-09 | 2023-04-18 | 东南大学 | Road confluence area roadside real-time accurate induction method for commercial vehicles |
| CN115352443B (en) * | 2022-09-21 | 2024-02-09 | 山东科技大学 | Self-adaptive cruise control method and device based on by-pass vehicle cut-in recognition |
| CN116880218B (en) * | 2023-09-06 | 2023-12-19 | 浙江大学 | Robust driving strategy generation method and system based on driving style misunderstanding |
| CN116946162B (en) * | 2023-09-19 | 2023-12-15 | 东南大学 | Intelligent network combined commercial vehicle safe driving decision-making method considering road surface attachment condition |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108225364A (en) * | 2018-01-04 | 2018-06-29 | 吉林大学 | A kind of pilotless automobile driving task decision system and method |
| CN112622886A (en) * | 2020-12-20 | 2021-04-09 | 东南大学 | Anti-collision early warning method for heavy operation vehicle comprehensively considering front and rear obstacles |
| CN112633474A (en) * | 2020-12-20 | 2021-04-09 | 东南大学 | Backward collision avoidance driving decision method for heavy commercial vehicle |
| WO2021238136A1 (en) * | 2020-12-04 | 2021-12-02 | 肇庆学院 | Hmm-rf hybrid model-based heavy vehicle rollover warning method and system |
| CN113753026A (en) * | 2021-10-21 | 2021-12-07 | 东南大学 | Decision-making method for preventing rollover of large commercial vehicle by considering road adhesion condition |
| CN113954837A (en) * | 2021-11-06 | 2022-01-21 | 交通运输部公路科学研究所 | Deep learning-based lane change decision method for large-scale commercial vehicle |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11635758B2 (en) * | 2019-11-26 | 2023-04-25 | Nissan North America, Inc. | Risk aware executor with action set recommendations |
| US11433924B2 (en) * | 2020-04-30 | 2022-09-06 | Toyota Motor Engineering & Manufacturing North America, Inc. | System and method for controlling one or more vehicles with one or more controlled vehicles |
-
2022
- 2022-02-21 CN CN202210157766.5A patent/CN114379540B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108225364A (en) * | 2018-01-04 | 2018-06-29 | 吉林大学 | A kind of pilotless automobile driving task decision system and method |
| WO2021238136A1 (en) * | 2020-12-04 | 2021-12-02 | 肇庆学院 | Hmm-rf hybrid model-based heavy vehicle rollover warning method and system |
| CN112622886A (en) * | 2020-12-20 | 2021-04-09 | 东南大学 | Anti-collision early warning method for heavy operation vehicle comprehensively considering front and rear obstacles |
| CN112633474A (en) * | 2020-12-20 | 2021-04-09 | 东南大学 | Backward collision avoidance driving decision method for heavy commercial vehicle |
| CN113753026A (en) * | 2021-10-21 | 2021-12-07 | 东南大学 | Decision-making method for preventing rollover of large commercial vehicle by considering road adhesion condition |
| CN113954837A (en) * | 2021-11-06 | 2022-01-21 | 交通运输部公路科学研究所 | Deep learning-based lane change decision method for large-scale commercial vehicle |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114379540A (en) | 2022-04-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114379540B (en) | Rollover-prevention driving decision method for large-sized operation vehicle considering influence of front obstacle | |
| CN109669461B (en) | Decision-making system for automatically driving vehicle under complex working condition and track planning method thereof | |
| CN112633474B (en) | Backward collision avoidance driving decision method for heavy commercial vehicle | |
| CN112622886B (en) | Anti-collision early warning method for heavy operation vehicle comprehensively considering front and rear obstacles | |
| US11648946B2 (en) | Intelligent vehicle platoon lane change performance evaluation method | |
| CN108657189B (en) | Automatic driving steering system based on BP neural network and safe distance lane change working condition and control method thereof | |
| CN112580148B (en) | Heavy-duty operation vehicle rollover prevention driving decision method based on deep reinforcement learning | |
| Tsugawa | An overview on an automated truck platoon within the energy ITS project | |
| CN110281893B (en) | Emergency braking system and method and semitrailer | |
| CN113753026B (en) | Decision-making method for preventing rollover of large commercial vehicle by considering road adhesion condition | |
| CN109131312B (en) | ACC/ESC integrated control system and method for intelligent electric vehicle | |
| CN111354225B (en) | Expressway vehicle accident risk assessment and early warning intervention method | |
| CN109664881A (en) | A kind of emergency turn auxiliary driving method, system and electronic equipment | |
| CN104527644A (en) | Self-adaption cruise system and method | |
| CN108717268A (en) | Automatic Pilot minimum time maneuver control system and its control method based on optimum control and safe distance | |
| CN113071482B (en) | Intelligent vehicle anti-collision method based on passing time interval | |
| CN114407931A (en) | Decision-making method for safe driving of highly-humanoid automatic driving commercial vehicle | |
| CN204452443U (en) | A kind of self-adaption cruise system | |
| CN116499486A (en) | Complex off-road environment path planning method and system and electronic equipment | |
| CN113071490A (en) | Highway truck formation system | |
| CN116039672A (en) | Data-model back-end fusion method for automatic driving decision control function | |
| Kanellakopoulos et al. | Commercial trucks and buses in automated highway systems | |
| Huang et al. | Dynamics performance of long combination vehicles with active control systems | |
| CN116472505A (en) | Method for reducing high-speed deviation of multi-trailer heavy vehicle | |
| CN113985883B (en) | Control system based on heterogeneous truck queue energy conservation, safety and cargo comfort |
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 |