CN110488802A - A kind of automatic driving vehicle dynamic behaviour decision-making technique netted under connection environment - Google Patents
A kind of automatic driving vehicle dynamic behaviour decision-making technique netted under connection environment Download PDFInfo
- Publication number
- CN110488802A CN110488802A CN201910774762.XA CN201910774762A CN110488802A CN 110488802 A CN110488802 A CN 110488802A CN 201910774762 A CN201910774762 A CN 201910774762A CN 110488802 A CN110488802 A CN 110488802A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- risk
- decision
- surrounding
- area
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000009471 action Effects 0.000 claims abstract description 50
- 230000006870 function Effects 0.000 claims description 65
- 230000001133 acceleration Effects 0.000 claims description 22
- 238000004891 communication Methods 0.000 claims description 14
- 230000006855 networking Effects 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 8
- 230000006399 behavior Effects 0.000 description 42
- 230000003542 behavioural effect Effects 0.000 description 13
- 230000008859 change Effects 0.000 description 9
- 238000011156 evaluation Methods 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 7
- 238000005259 measurement Methods 0.000 description 7
- 230000033228 biological regulation Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000008447 perception Effects 0.000 description 4
- 230000007613 environmental effect Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 206010039203 Road traffic accident Diseases 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003304 psychophysiological effect Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012502 risk assessment Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000013517 stratification Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/0055—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots with safety arrangements
- G05D1/0061—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots with safety arrangements for transition from automatic pilot to manual pilot and vice versa
Landscapes
- Engineering & Computer Science (AREA)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses the automatic driving vehicle dynamic behaviour decision-making techniques under a kind of net connection environment.The described method includes: step S1, from vehicle in the case where V2X nets connection environment, surrounding road user obtains surrounding enviroment information；And centered on from vehicle mass center, region division is carried out with different radiuses, estimates risk zones；Step S2 based on surrounding road user's surrounding enviroment information and estimates risk zones, carries out first stage behaviour decision making, is determined as the possible action set for guaranteeing to take from vehicle traffic safety；Step S3 carries out second stage behaviour decision making: considering non-safety constraint condition, from the possible action set, the movement that optimum choice finally executes carries out driving behavior decision.
Description
Technical Field
The invention relates to the field of automatic driving, in particular to a dynamic behavior decision method for an automatic driving vehicle in an internet environment.
Background
The intelligent traffic development brings convenience, but also causes potential traffic accidents. Therefore, safe driving of vehicles becomes a key factor of modern intelligent traffic systems. Although active safety systems and passive safety systems are becoming mature and applied to avoid vehicle collisions and minimize the impact of accidents, there is still a strong need to reduce the incidence of traffic accidents and improve the level of safety of vehicles. In recent years, autonomous vehicles have attracted considerable attention from the automotive industry due to their potential use in collision avoidance. However, automatic driving to achieve the goal of "zero accidents on the road" remains a complex task.
At present, the automatic driving vehicle can realize a plurality of basic functions due to various driving auxiliary systems, such as functions of lane changing auxiliary, self-adaptive cruise driving, early warning and the like. However, existing autonomous vehicle behavior decision algorithms cannot perform accurate risk assessment of potential risks on roads, i.e., these autonomous vehicle algorithms cannot identify which action is safer in an emergency, and cannot make effective and safe behavior decisions. In order to make an autonomous vehicle more intelligent, existing decision-making systems need to recognize and understand the comprehensive information of the driver-vehicle-road traffic environment, and an autonomous vehicle decision-making algorithm should consider the comprehensive driving risk and rank the severity of the driving risk by a driving risk index. Especially for complex road environments, the automatic driving automobile needs to adapt to a mixed traffic form, and meanwhile, the traffic efficiency is really improved and accidents are reduced.
The existing planning decision technology can be mainly divided into specific technical links such as task planning, route planning, behavior planning and motion planning. Typical vehicle decision planning involves different types of objects on the road. Some current research work considers the different operation of autonomous vehicles as separate driving modes and only in one of these modes, a.furda and l.vlacic also propose hierarchical decision methods to decide behavioral decision switches. However, these approaches typically rely on predefined strategies that are laborious to design and prone to be unreliable in emergency situations. Meanwhile, CVX and Ipopt also achieve fast and reliable path generation. The method can realize a complete track from initial configuration to target configuration and complete a decision planning process, but the method usually does not consider the dynamic property and the randomness of a driving environment and cannot accurately estimate risks, so that a safety obstacle avoidance decision is realized.
In the face of a complex and changeable operating environment, the automatic driving vehicle needs to be capable of accurately judging the current environmental risk value so as to carry out effective behavior decision. The limitation of the existing decision method causes the intelligent level of the intelligent vehicle to be limited, so that a decision method and a device for the dynamic behavior of the automatic driving vehicle in the internet environment need to be developed.
Disclosure of Invention
The invention aims to provide a method and a device for deciding the dynamic behavior of an automatic driving vehicle in a network connection environment, which can enable the automatic driving vehicle to effectively detect and evaluate the dangerous condition of the current traffic condition in a dynamic complex environment and make an effective behavior decision so that the automatic driving vehicle can avoid colliding with other road users.
In order to achieve the above object, the present invention provides a method for determining dynamic behavior of an autonomous vehicle in an internet environment, wherein the autonomous vehicle is an autonomous vehicle, and the method comprises the following steps S1-S3.
Step S1, the surrounding road users acquire surrounding environment information under the V2X networking environment; and carrying out area division by taking the center of mass of the self-vehicle as the center and different radiuses to estimate a risk area.
And step S2, performing a first-stage action decision based on the surrounding environment information of the surrounding road users and the estimated risk area, and determining a feasible action set which can be taken for ensuring the driving safety of the vehicle.
The first stage of behavioral decision is to make a safe and feasible decision in a specific scene. With respect to feasibility, decisions on safe driving maneuvers can be influenced by the environment surrounding the autonomous vehicle and by route planning instructions. A driving maneuver is defined as feasible if it can be safely performed under certain traffic conditions and complies with road traffic regulations. To ensure safety, it is assumed that the autonomous vehicle will always comply with traffic regulations, while satisfying a safety hard constraint function. In the first layer of decision, there may be a variety of driving maneuvers available (e.g., passing a stop or waiting for it to continue driving) in any traffic situation.
Step S3, performing a second stage action decision: and considering a non-safety constraint condition, optimally selecting a finally executed action from the feasible action set, and making a driving behavior decision. Step S3 corresponds to a decision to select the most appropriate driving maneuver from the plurality of feasible decisions (sets of feasible actions) of the first stage. The second stage behavior decision selects and begins to perform a single driving maneuver (driving behavior) that is selected to best suit the particular traffic situation. Non-safety constraints are for example requirements or constraints with respect to efficiency, comfort or traffic flow.
The V2V (vehicle-to-vehicle) networking environment may be part of an existing V2X networking environment, enabling reliable information exchange between autonomous vehicles through communication techniques. The V2X networking environment further enables reliable information exchange between the autonomous vehicle and road infrastructure (V2I) and the like on this basis. The invention can effectively realize the cooperation among vehicles in the networking environment and improve the efficiency and the safety of automatically driving the vehicles.
The self-vehicle obtains information from the vehicle-mounted sensor system by using a perception algorithm, and effectively identifies relevant traffic characteristics, such as traffic signs, road marks, obstacles, pedestrians, vehicles and the like. Sensor technology for on-board sensor systems can provide accurate and reliable information about the vehicle environment under any light, weather and road conditions. Meanwhile, the information such as the position (longitude and latitude information), the height, the speed, the acceleration and the like of the vehicle is acquired and sent to the surrounding road users, the same information sent by the surrounding road users is received, and the distance and the position of the surrounding road users relative to the vehicle can be obtained through calculation.
The self vehicle stores the information as required after acquiring the information, and updates the information if necessary. Storing environmental priors, such as roads, intersections, and traffic signs; storing information provided by sensors, such as obstacles, traffic lanes, and perceived traffic signs; storing information obtained by communicating with other vehicles or a traffic management center; the prior information is continuously updated with information continuously obtained from the sensors and the communication assembly.
Preferably, in step S1, the risk area is estimated in the following manner:
defining the safe braking distance L by taking the center of mass of the bicycle as the centerriskThe circular area of radius is the risk area,
in the formula, viIs the current speed of the vehicle, amaxThe maximum value of the acceleration of the bicycle is L, and the length of the bicycle is L; if a certain surrounding road user is located in the risk area, the surrounding road user is defined as a risk road user;
defining the safety early warning distance L with the center of mass of the bicycle as the centerp-riskThe annular region after the risk region is removed for the circular region of radius is the potential risk region,
wherein, adecThe maximum value of the deceleration of the vehicle, if a certain surrounding road user is located in the potential risk area, the surrounding road user is defined as a potential risk road user;
define with self-vehicleCentroid as center, safety early warning distance Lp-riskThe area outside the circular area of the radius is a safe area, and if a certain surrounding road user is located outside the potential risk area or outside the communication range of the own vehicle, the surrounding road user is defined as a safe road user.
Preferably, step S2 includes the steps of:
step S21, calculating a risk degree C for the risk area and the surrounding road users within the potential risk area, wherein the calculation is performed for the surrounding road users within the risk area first, then for the surrounding road users within the potential risk area,
the risk degree C represents the probability of collision between the current state of the autonomous vehicle and the state of the surrounding road users, that is, the probability of collision or collision when the current state of the autonomous vehicle and the surrounding road users is kept unchanged. The probability is an estimate and is calculated by:
wherein t is the estimated predicted time to collision, t is the minimum of the two or more estimated predicted time to collision with each surrounding road user if there are two or more surrounding road users in the risk area and the potential risk area, and t is greater than t if there are no surrounding road users in the risk area and the potential risk areacThe setting value of (a) is set,
tcthe critical time for avoiding a collision may be predetermined as a set constant according to the braking performance of the vehicle, the emergency response speed, the road performance, and the like. For example, in one embodiment of the present invention, t is definedc＝4s。
When C is 0, judging that no traffic conflict exists in the state, and determining a risk metric value friskGo to zero, go to step S23;
when C is 1, it is determined that there is a potential traffic collision in this state, go to step S22,
step S22, calculating a risk metric frisk，
Step S23, according to the risk metric friskAnd selecting the action and determining a feasible action set.
Preferably, the estimated predicted time to collision t is calculated by the following equation,
t＝min{TTC,PET,TTB}
wherein,
Xiis the position of the vehicle at the self-parking position,
Xjis the position of the following surrounding road user,
viin order to obtain the current speed of the vehicle,
vjthe current speed of the other vehicle is the current speed,
Liin order to obtain the length of the bicycle,
PET is the time t for the vehicle to enter the conflict pointiTo the time t when another surrounding road user reaches the conflict pointjThe difference between the times of the two phases,
PET＝t＝|ti-tj|
TTB is used for evaluating the forward area, is suitable for the scene that the own vehicle is behind and the other vehicles are in front,
Xiis the position of the vehicle at the self-parking position,
Xjis the location of the other vehicle followed by,
viin order to obtain the current speed of the vehicle,
Liis the length of the bicycle.
In one embodiment, the estimated predicted time to collision t is calculated in the following manner.
On the one hand, when a scene can be distinguished, only the estimated predicted time-to-collision t for that scene is calculated, wherein for a straight-following scene, t is TTC; for an intersection scene, t ═ PET; for the scene of collision of the self vehicle in the rear-front direction, t is TTB; that is to say, when a scene can be distinguished, only the risk degree function for the scene needs to be calculated, specifically, the TTC is mainly used for a straight-road following scene, the PET is suitable for an intersection scene, and the TTB is suitable for a rear-front collision scene of a self-vehicle. When calculating the risk measurement, the TTC index is selected for evaluation, and other applicable indexes are selected for the TTC scene which is not applicable. And when the evaluation results of the three indexes are different, selecting the index with the highest risk for decision output. And further analyzing and intervening decision-making according to the risk degree of multi-target output.
PET can capture primarily the impact of certain other intersection characteristics on safety, since the inclusion of other intersection characteristics (e.g., line of sight, grade, and other parameters) only marginally affects prediction ability, and thus the primary applicable scenario is intersection. TTC can be applied to different types of collisions, such as rear, front and right angle collisions, but it is more accurate in straight-track scenes. TTB is mainly used to evaluate the forward zone, i.e. measurements are not used in the rear zone, applicable to the scenario where the own vehicle is in front of the other vehicles.
On the other hand, when the scene is complex and difficult to distinguish, the three indexes are calculated to take the minimum value,
t＝min{TTC,PET,TTB}，
wherein,
Xiis the position of the vehicle at the self-parking position,
Xjis the position of the following surrounding road user,
viin order to obtain the current speed of the vehicle,
vjthe current speed of the other vehicle is the current speed,
Liin order to obtain the length of the bicycle,
PET is the time t for the vehicle to enter the conflict pointiTo the time t when another surrounding road user reaches the conflict pointjThe difference between the times of the two phases,
PET＝t＝|ti-tj|
TTB is used for evaluating the forward area, is suitable for the scene that the own vehicle is behind and the other vehicles are in front,
Xiis the position of the vehicle at the self-parking position,
Xjis the location of the other vehicle followed by,
viin order to obtain the current speed of the vehicle,
Liis the length of the bicycle.
In different traffic scenarios, if there are multiple other road users in the same scenario, the estimated projected time-to-collision for each road user is evaluated separately, and then a "valid" or "equivalent" estimated projected time-to-collision risk t is determined by using a multi-objective threat assessment algorithm. The multi-target threat assessment algorithm adopts a minimum value taking algorithm or a weighted minimum value taking algorithm, for example.
The weighted minimum algorithm is, for example: assigning a weighting factor a, a >0.5 to the smallest estimated predicted time-to-collision t 1; assigning a weighting factor a (1-a) to the second smallest estimated predicted time-to-collision t 2; the third smallest estimated predicted time-to-collision t2 is given a weighting coefficient a (1-a-a (1-a)), etc. a is for example 0.6 or 2/3.
The decision framework provided by the invention is suitable for multi-target scenes, and further analysis and decision intervention are carried out according to the risk degree of multi-target output.
Preferably, in step S2, the risk metric value f is calculated by the following equationrisk，
Or
Preferably, at the second step of S3In the decision of two-stage behaviors, any decision attribute influencing safety is not included, and an efficient soft constraint function f is consideredeComfort soft constraint function fcAnd a traffic flow soft constraint function ftAnd carrying out optimal decision.
And the second stage behavior decision is the second decision stage. And finding the optimal decision method in the second decision stage. The decision to select the most appropriate driving maneuver from the plurality of possible decisions of the first floor. This phase selects and begins to perform a single driving maneuver that is selected to best suit a particular traffic situation. Since only those driving maneuvers that are selected to be feasible (and therefore safe) are considered at this stage, it is not safety critical that this stage does not include any decision-making attributes that affect safety. The second decision phase is mainly determined by an efficient soft constraint function feComfort soft constraint function fcTraffic flow soft constraint function ft。
Preferably, the efficient soft constraint function feIs defined as:
wherein, t0Is the initial departure time of the vehicle, tfWhen the vehicle reaches the destination, s (t) is a path traveled by the vehicle from the starting point to the destination, and (t) is a vehicle speed.
Preferably, the comfort soft constraint function fcIs defined as:
wherein a is the acceleration of the bicycle, alatFor lateral acceleration, alonIs the longitudinal acceleration.
The comfort is reflected by the performance index of the vehicle, and the invention mainly aims at the change of the psycho-physiological comfort of passengers caused by the horizontal vibration of the mechanical structure and the assembly manufacturing of the vehicle caused by the maneuverability of the decision of automatically driving the vehicle, wherein the handling behaviors of rapid acceleration, rapid deceleration and the like can obviously impact the passengers.
Preferably, the traffic flow soft constraint function ftIs defined as:
minft＝α(vave-vder)2+β(dave-dder)2
wherein,
vaveto determine the average speed level of the surrounding traffic flow before decision,
vderto determine the desired average speed level of the surrounding traffic flow,
daveto determine the average inter-vehicle distance of the surrounding traffic flow before decision,
dderto determine the desired average inter-vehicle distance for the surrounding traffic flow,
alpha and beta are weight coefficients which are both larger than 0 and smaller than 1.
So that the disturbance to the surrounding vehicle is minimized. The disturbing effect of the change of the state of the own vehicle on the dynamic characteristics of the surrounding traffic flow is reflected in the change of the average speed and the distance level of other vehicles from the expected speed and the expected distance level of other vehicles.
The potential influence brought by the decision-making behavior of the automatic driving vehicle, namely the disturbance effect of the state change of the automatic driving vehicle on the dynamic characteristics of the surrounding traffic flow is particularly reflected in the influence on the traffic capacity of the surrounding roads and the stability and smoothness of the traffic flow. Therefore, a traffic flow soft constraint function f is definedtWhen the decision of the behavior of the automatic driving vehicle is judged, the influence of the behavior on the traffic flow, which is reflected in the reduction of the speed or the interruption of the acceleration driving, needs to be measured.
In the internet environment, the surrounding traffic flow range is defined as vehicles within the communication range of the own vehicle.
Preferably, in the second stage of the behavior decision of step S3, the cost function J is defined as follows:
w1、w2、w3are weight coefficients, all are greater than 0 and less than 1, and w1+w2+w3＝1
Wherein f ise0、fc0、ft0Respectively representing the safety, high efficiency, comfort and traffic flow functions after the self vehicle is supposed to continuously execute according to the state before decision.
And judging whether the decision is reasonable or not, namely comparing the cost function J after the decision is made. The minimum value of the cost function J corresponds to the optimal solution.
The method of the invention estimates the risk area according to the traffic risk function through information interaction between vehicles, carries out behavior decision in stages, improves decision speed, efficiency and safety, and can help realize safe, efficient and comfortable destination arrival by controlling the action of automatically driving vehicles.
Drawings
Fig. 1 is a schematic flow chart of a method for determining dynamic behavior of an autonomous vehicle according to an embodiment of the present invention.
FIG. 2 is a schematic block diagram of a control system that employs a method for dynamic behavior decision-making for an autonomous vehicle provided by an embodiment of the invention;
fig. 3 is a schematic diagram of risk classification areas according to an embodiment of the present invention.
Fig. 4 is a diagram illustrating a first stage of behavioral decision in the present invention.
FIG. 5 is a diagram illustrating a first stage behavioral decision and a second stage behavioral decision in the present invention.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
Referring to fig. 1, the method for determining the dynamic behavior of the autonomous vehicle in the internet environment according to the embodiment of the present invention includes the following steps S1-S3.
Step S1, the surrounding road users acquire surrounding environment information under the V2X networking environment; and carrying out area division by taking the center of mass of the self-vehicle as the center and different radiuses to estimate a risk area. The self vehicle is an automatic driving vehicle. The surrounding environment information includes, for example, position and speed information of a road user such as a surrounding vehicle, a rider, a pedestrian, and an obstacle. The estimated risk area is equivalent to the preprocessing of a risk prediction link in fig. 4, and the risk area is determined.
Further, in decision stage 1 (i.e., the first stage behavior decision), for example, as shown in fig. 2, in the risk pre-judging section, a risk road user, a potential risk road user and a safe road user are determined by judging whether the road user is in a risk area. In one embodiment, the distance between the other vehicle and the own vehicle is defined as the distance between the head of the other vehicle and the centroid of the own vehicle in the connecting line direction.
Further, in step S1, pre-calculations or determinations may be made for the first stage behavior decision in step S2, such as determining potential conflicts, estimating time of conflicts, probability of conflicts, etc., by the traffic risk stratification assessment module (see fig. 2). The above determination, estimation, and the like may be calculated in step S2. For example, the above estimation and determination are performed in the collision detection section shown in fig. 4.
The Vehicle with the V2X (Vehicle to event) function generally refers to a Vehicle that is equipped with advanced Vehicle-mounted sensors, controllers, actuators and other devices (i.e., an internet of vehicles network sensing module in fig. 2), integrates modern communication and network technologies, realizes exchange and sharing of Vehicle and X (people, vehicles, roads, backgrounds and the like) intelligent information, has functions of complex environment sensing, intelligent decision making, cooperative control, execution and the like, can realize safe, comfortable, energy-saving and efficient driving, and can finally replace people to operate. Autonomous vehicles include autonomous vehicles of respective autonomous driving category classes.
Compared with a camera or a laser radar which is commonly used in the automatic driving technology, the V2X has a wider application range, has the information acquisition capability of breaking through visual dead corners and crossing shelters, can share real-time driving state information with other vehicles and facilities, and can generate prediction information through a research and judgment algorithm. In addition, V2X is the only vehicle sensing technology that is not affected by weather conditions, and does not affect normal operation of the vehicle regardless of rain, fog or strong light.
In addition, in addition to the functions of traditional intelligent automobile information exchange sharing and environment perception, V2X also emphasizes the functions of "intelligent decision making", "cooperative control and execution", which are based on a powerful background data analysis, decision making, and scheduling service system. In addition, to realize automatic driving, the vehicle must have a sensing system to observe the surrounding environment like a human, so the V2X technology belongs to a sensing means for automatic driving in addition to the sensor.
The V2V (vehicle-to-vehicle) networking environment may be part of an existing V2X networking environment, enabling reliable information exchange between autonomous vehicles through communication techniques. The V2X networking environment further enables reliable information exchange between the autonomous vehicle and the road infrastructure (V2I) on this basis. The invention can effectively realize the cooperation among vehicles in the networking environment and improve the efficiency and the safety of automatically driving the vehicles.
The self-vehicle obtains information from the vehicle-mounted sensor system by using a perception algorithm, and effectively identifies relevant traffic characteristics, such as traffic signs, road marks, obstacles, pedestrians, vehicles and the like. Sensor technology for on-board sensor systems can provide accurate and reliable information about the vehicle environment under any light, weather and road conditions. Meanwhile, the information such as the position (longitude and latitude information), the height, the speed, the acceleration and the like of the vehicle is acquired and sent to the surrounding road users, the same information sent by the surrounding road users is received, and the distance and the position of the surrounding road users relative to the vehicle can be obtained through calculation.
The self vehicle stores the information as required after acquiring the information, and updates the information if necessary. Storing environmental priors, such as roads, intersections, and traffic signs; storing information provided by sensors, such as obstacles, traffic lanes, and perceived traffic signs; storing information obtained by communicating with other vehicles or a traffic management center; the prior information is continuously updated with information continuously obtained from the sensors and the communication assembly.
Steps S2 and S3 are performed in the multi-tiered behavioral decision module.
Step S2, based on the surrounding environment information of the users on the surrounding roads and the estimated risk area, a first-stage action decision (decision stage 1 in fig. 2) is performed to determine a set of feasible actions that can be taken to ensure the safety of the vehicle. The first stage behavioral decision meets the requirements of a safety hard constraint function. The result is a set of feasible actions (i.e., all of the feasible decision sets in FIG. 2).
As shown in fig. 4, the first stage of behavioral decision generally includes the following steps: a risk pre-judging link, a conflict retrieval link, a risk measurement link and an action selection link. The risk pre-judging link mainly judges the risk road users. The collision detection link is mainly used for preliminarily judging whether a collision (collision) occurs. The risk measure is mainly a calculated risk measure value. The risk metric value may be calculated in any suitable manner. In addition to the risk metric values described in detail below,
the first stage of behavioral decision is to make a safe and feasible decision in a specific scene. With respect to feasibility, decisions on safe driving maneuvers can be influenced by the environment surrounding the autonomous vehicle and by route planning instructions. A driving maneuver is defined as feasible if it can be safely performed under certain traffic conditions and complies with road traffic regulations. To ensure safety, it is assumed that the autonomous vehicle will always comply with traffic regulations, while satisfying a safety hard constraint function. In the first layer of decision, there may be a variety of driving maneuvers available (e.g., passing a stop or waiting for it to continue driving) in any traffic situation.
Step S3, performing a second stage behavior decision (i.e., decision stage 2 in fig. 2): and (5) performing second-stage behavior decision: and considering a non-safety constraint condition, optimally selecting a finally executed action from the feasible action set, and making a driving behavior decision. Step S3 corresponds to a decision to select the most appropriate driving maneuver from the plurality of feasible decisions (sets of feasible actions) of the first stage. The second stage behavior decision selects and begins to perform a single driving maneuver (driving behavior) that is selected to best suit the particular traffic situation. Non-safety constraints are for example requirements or constraints with respect to efficiency, comfort or traffic flow. That is, the second stage action decision is made primarily based on an efficient, comfortable, traffic flow constraint function. The result is to obtain or make optimal behavioral decisions. Referring to fig. 5, the second stage behavioral decision makes an optimal behavioral decision based on the set of selectable actions output from the first stage behavioral decision.
According to some embodiments of the invention, in step S1, the risk areas are estimated in the following manner:
defining the safe braking distance L by taking the center of mass of the bicycle as the centerriskThe circular area of radius is the risk area,
in the formula, viIs the current speed of the vehicle, amaxThe maximum value of the acceleration of the bicycle is L, and the length of the bicycle is L; if a certain surrounding road user is located in the risk area, the surrounding road user is defined as a risk road user;
defining the safety early warning distance L with the center of mass of the bicycle as the centerp-riskThe annular region after the risk region is removed for the circular region of radius is the potential risk region,
wherein, adecThe maximum value of the deceleration of the vehicle, if a certain surrounding road user is located in the potential risk area, the surrounding road user is defined as a potential risk road user;
defining the safety early warning distance L with the center of mass of the bicycle as the centerp-riskThe area outside the circular area of the radius is a safe area, and if a certain surrounding road user is located outside the potential risk area or outside the communication range of the own vehicle, the surrounding road user is defined as a safe road user. Namely that the area where the safety road users are located is at the radius Lsafe＝min{Lcom,Lp-riskArea outside. Wherein at a communication radius LcomThe round range means that the vehicle can be placed in the designated area under the networking environmentTo observe and acquire the range of the user information of the surrounding road.
Specifically, as shown in fig. 3, the present invention uses the center of mass of the vehicle as the center and the radius as the communication radius length LcomThe circle of (2) is divided into regions, and specifically, the regions are divided into risk road users, potential risk road users and safety road users according to risk grades. For an observed vehicle, three distance metrics are defined that correlate to the three risk levels described above, and if the range exceeds a threshold, the risk metric for the observed vehicle changes accordingly. Wherein, as shown in FIG. 3, the boundary between the risk road users and the potential road users is defined as the safe braking distance LriskIs a circle with a radius, and the inner part of the circular area is a risk area. Defining the boundary between the potential risk road user and the safe road user as the safe early warning distance Lp-riskWhen in a circle of radius (L)risk,Lp-risk) Within the annular region in between are potential risk regions. When the distance between the vehicle and the conflict point is larger than the safety warning distance or the other vehicle is not in the communication range of the vehicle, the vehicle is defined as a safety road user. Namely that the area where the safety road users are located is at the radius Lsafe＝min{Lcom,Lp-riskArea outside. Wherein at a communication radius LcomThe round range refers to a range in which the user information of the surrounding road can be observed and acquired in the designated area of the vehicle under the networking environment.
In one embodiment, step S2 includes the steps of:
step S21, calculating a risk degree C for the risk area and the surrounding road users within the potential risk area, wherein the calculation is performed for the surrounding road users within the risk area first, then for the surrounding road users within the potential risk area,
the risk degree C represents the probability of collision between the current state of the autonomous vehicle and the state of the surrounding road users, that is, the probability of collision or collision when the current state of the autonomous vehicle and the surrounding road users is kept unchanged. The probability is an estimate and is calculated by:
wherein t is the estimated predicted time to collision, t is the minimum of the two or more estimated predicted time to collision with each surrounding road user if there are two or more surrounding road users in the risk area and the potential risk area, and t is greater than t if there are no surrounding road users in the risk area and the potential risk areacThe setting value of (a) is set,
tcthe critical time for avoiding a collision may be predetermined as a set constant according to the braking performance of the vehicle, the emergency response speed, the road performance, and the like. For example, in one embodiment of the present invention, t is definedc＝4s。
When C is 0, judging that no traffic conflict exists in the state, and determining a risk metric value friskGo to zero and go to step S23.
When C is 1, it is determined that there is a potential traffic collision in this state, and the process proceeds to step S22.
Step S22, calculating a risk metric friskThe value of risk metric friskThe calculation may be performed in any suitable way. For ease of evaluation, the risk metric friskSet to a number between 0-1.
Step S23, according to the risk metric friskAnd selecting the action and determining a feasible action set.
It is noted that the total set of actions that the vehicle is capable of performing is different in different specific scenarios. See, for example, table 1 for a total set of actions in each scenario.
TABLE 1 general set of actions for autonomous vehicles in different traffic scenarios
| Action sequence/traffic scene | Case 1 | Case 2 | Case 3 | Case 4 | Case p |
| Accelerated passage | √ | √ | √ | — | |
| Directly through | √ | √ | — | ||
| Speed reduction | √ | √ | √ | — | |
| Waiting for parking | √ | √ | — | ||
| Follow in turn | √ | √ | — | ||
| Turning passage | √ | √ | — | ||
| Action q | — | — | — | — | — |
For some classical scenarios, the optional results of decision stage 1 are listed in table 1. Further description of Table 1 is as follows:
case 1: a free-driving scenario. Autonomous vehicles travel along roads relatively far from an upcoming intersection. In this case, the risk may be output based on a first level assessment and a second level decision may be made to provide three possible speed execution options (i.e., acceleration, raw speed, deceleration).
Case 2: an intersection scene. Autonomous vehicles are approaching an intersection with other dynamic road users (e.g., a pedestrian crossing the road, a cyclist, other vehicles, etc.). In this case, the risk may be exported based on a first level assessment and then a second level decision may be made to provide four operational execution options (i.e., stop waiting, follow-up, direct pass, turn pass) and three speed execution options (i.e., acceleration, speed, deceleration) that are feasible.
Case 3: and (5) following scenes. An autonomous vehicle is following another vehicle. In this case, the risk may be output in judgment based on the first-tier evaluation, and then the second-tier decision is made, and it is possible to determine that two operation choices (i.e., follow-up, turn-through) and the speed execution choice (i.e., acceleration, deceleration) are possible that the driving operation is possible.
Case 4: the obstacle blocks the scene. Static obstacles (trees) are located in front of the autonomous vehicle. The only feasible driving operation is an emergency stop.
Case p: refers to driving scenarios other than the 4 cases described above.
Action q: it means other than the above-mentioned 7 operations.
"-": it is shown that it is possible to determine whether it is feasible according to a specific scene and a specific action.
The first stage of behavior decision is to select a feasible action set (i.e. all feasible decision sets in fig. 2) from the action total set in table 1 under the premise of ensuring security. More specifically, in step S23, based on the risk metric value friskAnd selecting the action and determining a feasible action set.
Table 2 illustrates different risk metrics f using an intersection scenario as an exampleriskThe corresponding set of feasible actions.
TABLE 2 alternative action selection mapping table for autonomous vehicles under different risk metric values
| Action selection/risk metric | 0 | (0,0.4] | (0.4,0.7] | (0.7,1] |
| Accelerated passage | √ | |||
| Directly through | √ | |||
| Slow down and go slowly | √ | √ | ||
| Waiting for parking | √ | √ | √ | |
| Follow in turn | √ | √ | ||
| Turning passage | √ | √ | √ | √ |
In the specific embodiment of the invention, the specific decision process can be divided into the steps of judging the security attribute for state selection under the drive of external conditions, namely, determining through a security hard constraint function; and then taking decision actions such as acceleration and deceleration driving, turning lane changing, speed keeping, parking waiting and the like according to specific scenes.
For example, in a specific case of considering efficiency (high efficiency), comfort and traffic flow, through optimization calculation, the final decision takes different actions for different specific driving situations.
TABLE 3 actions determined based on different efficiency, comfort and traffic flow requirements
In some embodiments of the present invention, in decision stage 1, a safety hard constraint function (risk metric function) is introduced to make a safety feasible decision under a specific scenario.
In an alternative embodiment of the invention, the estimated predicted time to collision, t, is calculated by the following equation,
t＝min{TTC,PET,TTB}
wherein,
Xiis the position of the vehicle at the self-parking position,
Xjis the position of the following surrounding road user,
viin order to obtain the current speed of the vehicle,
vjthe current speed of the other vehicle is the current speed,
Liin order to obtain the length of the bicycle,
PET is the time t for the vehicle to enter the conflict pointiTo the time t when another surrounding road user reaches the conflict pointjThe difference between the times of the two phases,
PET＝t＝|ti-tj|
TTB is used for evaluating the forward area, is suitable for the scene that the own vehicle is behind and the other vehicles are in front,
Xiis the position of the vehicle at the self-parking position,
Xjis the location of the other vehicle followed by,
viin order to obtain the current speed of the vehicle,
Liis the length of the bicycle.
In one embodiment, the estimated predicted time to collision t is calculated in the following manner.
On the one hand, when a scene can be distinguished, only the estimated predicted time-to-collision t for that scene is calculated, wherein for a straight-following scene, t is TTC; for an intersection scene, t ═ PET; for the scene of collision of the self vehicle in the rear-front direction, t is TTB; that is to say, when a scene can be distinguished, only the risk degree function for the scene needs to be calculated, specifically, the TTC is mainly used for a straight-road following scene, the PET is suitable for an intersection scene, and the TTB is suitable for a rear-front collision scene of a self-vehicle. When calculating the risk measurement, the TTC index is selected for evaluation, and other applicable indexes are selected for the TTC scene which is not applicable. And when the evaluation results of the three indexes are different, selecting the index with the highest risk for decision output. And further analyzing and intervening decision-making according to the risk degree of multi-target output.
PET can capture primarily the impact of certain other intersection characteristics on safety, since the inclusion of other intersection characteristics (e.g., line of sight, grade, and other parameters) only marginally affects prediction ability, and thus the primary applicable scenario is intersection. TTC can be applied to different types of collisions, such as rear, front and right angle collisions, but it is more accurate in straight-track scenes. TTB is mainly used to evaluate the forward zone, i.e. measurements are not used in the rear zone, applicable to the scenario where the own vehicle is in front of the other vehicles.
On the other hand, when the scene is complex and difficult to distinguish, the three indexes are calculated to take the minimum value,
t＝min{TTC,PET,TTB}，
wherein,
Xiis the position of the vehicle at the self-parking position,
Xjis the position of the following surrounding road user,
viin order to obtain the current speed of the vehicle,
vjthe current speed of the other vehicle is the current speed,
Liin order to obtain the length of the bicycle,
PET is the time t for the vehicle to enter the conflict pointiTo the time t when another surrounding road user reaches the conflict pointjThe difference between the times of the two phases,
PET＝t＝|ti-tj|
TTB is used for evaluating the forward area, is suitable for the scene that the own vehicle is behind and the other vehicles are in front,
Xiis the position of the vehicle at the self-parking position,
Xjis the location of the other vehicle followed by,
viin order to obtain the current speed of the vehicle,
Liis the length of the bicycle.
In different traffic scenarios, if there are multiple other road users in the same scenario, the estimated projected time-to-collision for each road user is evaluated separately, and then a "valid" or "equivalent" estimated projected time-to-collision risk t is determined by using a multi-objective threat assessment algorithm. The multi-target threat assessment algorithm adopts a minimum value taking algorithm or a weighted minimum value taking algorithm, for example.
The weighted minimum algorithm is, for example: assigning a weighting factor a, a >0.5 to the smallest estimated predicted time-to-collision t 1; assigning a weighting factor a (1-a) to the second smallest estimated predicted time-to-collision t 2; the third smallest estimated predicted time-to-collision t2 is given a weighting coefficient a (1-a-a (1-a)), etc. a is for example 0.6 or 2/3.
The decision framework provided by the invention is suitable for multi-target scenes, and further analysis and decision intervention are carried out according to the risk degree of multi-target output.
The risk level of the vehicle is more accurately evaluated by calculating the risk metric value. In particular, the risk measure of an observed vehicle depends on the distance between the observed vehicle and the autonomous vehicle and the relative speeds of the two vehicles. To assess the potential threat of collision between an observed vehicle and an autonomous vehicle, TTC is used as a common threat measure because it is time-based and involves spatial proximity and speed differences. However, in some highway situations, TTC alone is not sufficient. For example, consider the case: vehicles observed in lanes adjacent to the autonomous vehicle travel in close proximity to the autonomous vehicle and at similar speeds. This will result in a large TTC value, thereby assessing the situation as low risk (if only TTC is used). In practice, however, the example case is very dangerous in nature; therefore, in this case, the lane change maneuver should not be performed (to avoid the collision). Further, the TTC does not take into account exceptional situations, such as a situation where the vehicle observed in front suddenly stops for some reason. Therefore, to supplement TTC, two other evaluation methods were also adopted for situation evaluation: (1) PET, which describes the difference in time between two vehicles entering the conflict point, (2) TTB, which is defined as the remaining time that emergency braking should be applied to prevent a sudden stop or a maneuver collision. According to the definition of TTB, the measurement of TTB is not used in the rear area. Therefore, the different risk metric values are used in different stages of defining the subareas to evaluate the risk levels of the vehicles in different areas.
The determination of whether a conflict exists is typically measured by estimating the criticality of the traffic condition. Up to now, various safety indexes have been developed, such as Time To Collision (TTC), time after intrusion (PET), Unsafe Density (UD), deceleration rate to avoid collision (DRAC), Proportion of Stopping Distance (PSD), Gap Time (GT), integrated time measurement (CTM), rear-end collision probability (RECP), etc., of which one of the most widely used indexes is Time To Collision (TTC). Its definition is the time remaining before a collision with two vehicles traveling on the same route at the initial speed. However, when the TTC determines a dangerous situation, the two vehicles are generally close to the collision point, and thus it is not well reflected whether there is a potential collision. The invention is measured using the criterion PET (nonstencime) measured by the time two vehicles pass through the same position. One observable measure that allows consistency between the observer and the location is time-after-invasion (PET). PET is the difference between when a first vehicle ends a collision zone and when a second vehicle enters the collision zone. PET requires only two time stamps for computation, which has the advantage of a well-defined boundary to distinguish between crash and non-crash events. A PET value of 0 indicates a collision, while a non-zero PET value indicates a collision is close. Although it does not describe the initial phase of the collision, nor the action taken by the associated driver, it shows the results of the final phase and provides a measure of relative proximity to the collision. Current research evaluates the effectiveness of PET as an alternative safety measure to prevent objections from occurring through vehicle collisions.
In an alternative embodiment of the present invention, in step S2, the risk metric value f is calculated by the following equationrisk，
In this alternative embodiment, let t be 3s, frisk(4-3)/4-0.25. Let t be 1s, frisk(4-1)/4-0.75. Let t be 0.4s, frisk＝(4-0.4)/4＝0.9。
In another alternative embodiment of the present invention, in step S2, the risk metric value f is calculated by the following equationrisk，
In this further alternative embodiment, the risk metric value increases rapidly as the estimated time to collision decreases. Thus, the risk is further highlighted. Specifically, in this alternative embodiment, let t be 3s, frisk(4 × 4-3 × 3)/(4 × 4) ═ 0.44. Let t be 1s, frisk(4 × 4-1 × 1)/(4 × 4) ═ 0.94 assuming t is 0.4s, frisk＝(4*4-0.4*0.4)/(4*4)＝0.99。
Specifically, in the second-stage behavior decision of step S3, no decision attribute affecting security is included, but the efficient soft constraint function f is consideredeComfort soft constraint function fcAnd a traffic flow soft constraint function ftAnd carrying out optimal decision. It is also possible to consider only the other constraint functions, or only one or two of the above soft constraint functions.
And the second stage behavior decision is the second decision stage. And finding the optimal decision method in the second decision stage. The decision to select the most appropriate driving maneuver from the plurality of possible decisions of the first floor. This phase selects and begins to perform a single driving maneuver that is selected to best suit a particular traffic situation. Due to the fact thatOnly those driving manoeuvres that are selected to be feasible (and therefore safe) are considered at this stage, so it is not important that this stage includes safety, as it does not include any decision-making attributes that affect safety. The second decision phase is mainly determined by an efficient soft constraint function feComfort soft constraint function fcTraffic flow soft constraint function ft。
Preferably, the efficient soft constraint function feIs defined as:
wherein, t0Is the initial departure time of the vehicle, tfS (t) is the time when the vehicle reaches the destination, S (t) is the path from the starting point to the destination, and v (t) is the speed of the vehicle.
Preferably, the comfort soft constraint function fcIs defined as:
wherein a is the acceleration of the bicycle, alatFor lateral acceleration, alonIs the longitudinal acceleration.
The comfort is reflected by the performance index of the vehicle, and the invention mainly aims at the change of the psycho-physiological comfort of passengers caused by the horizontal vibration of the mechanical structure and the assembly manufacturing of the vehicle caused by the maneuverability of the decision of automatically driving the vehicle, wherein the handling behaviors of rapid acceleration, rapid deceleration and the like can obviously impact the passengers.
Preferably, the traffic flow soft constraint function ftIs defined as:
minft＝α(vave-vder)2+β(dave-dder)2
wherein,
vaveto determine the average speed level of the surrounding traffic flow before decision,
vderto determine the desired average speed level of the surrounding traffic flow,
daveto determine the average inter-vehicle distance of the surrounding traffic flow before decision,
dderto determine the desired average inter-vehicle distance for the surrounding traffic flow,
alpha and beta are weight coefficients which are both larger than 0 and smaller than 1.
So that the disturbance to the surrounding vehicle is minimized. The disturbing effect of the change of the state of the own vehicle on the dynamic characteristics of the surrounding traffic flow is reflected in the change of the average speed and the distance level of other vehicles from the expected speed and the expected distance level of other vehicles.
The potential influence brought by the decision-making behavior of the automatic driving vehicle, namely the disturbance effect of the state change of the automatic driving vehicle on the dynamic characteristics of the surrounding traffic flow is particularly reflected in the influence on the traffic capacity of the surrounding roads and the stability and smoothness of the traffic flow. Therefore, a traffic flow soft constraint function f is definedtWhen the decision of the behavior of the automatic driving vehicle is judged, the influence of the behavior on the traffic flow, which is reflected in the reduction of the speed or the interruption of the acceleration driving, needs to be measured.
In the internet environment, the surrounding traffic flow range is defined as vehicles within the communication range of the own vehicle.
Preferably, in the second stage of the behavior decision of step S3, the cost function J is defined as follows:
w1、w2、w3are weight coefficients, all are greater than 0 and less than 1, and w1+w2+w3＝1。w1、w2、w3The system can be preset, and can also be adjusted according to the requirements of the passengers. For example, if the current demand of the occupant is that the vehicle will arrive at the destination at the fastest speed, w1The settings are large, e.g. 0.5, 0.6, 0.8, or even 1. The system may also force the setting of w3Greater than or equal to a minimum value, such as 0.1.
Wherein f ise0、fc0、ft0Respectively representing the efficiency of the vehicle after the vehicle is supposed to continuously execute according to the state before decisionComfort, traffic flow function. It is noted that the efficient, comfortable, traffic flow function is not limited to the functions given above, but may also take any known form of corresponding constraint function. Moreover, embodiments employing corresponding efficient, comfortable, traffic flow restriction functions of known form, as well as embodiments including other non-safety restriction functions, are within the scope of the present invention.
And judging whether the decision is reasonable or not, namely comparing the cost function J after the decision is made. The minimum value of the cost function J corresponds to the optimal solution.
According to the method provided by the embodiment of the invention, through information interaction among vehicles, the risk area is estimated according to the traffic risk function, and behavior decision is carried out in stages, so that the decision speed, efficiency and safety are improved, and the aim of safely, efficiently and comfortably reaching the destination by controlling the action of automatically driving the vehicle can be realized.
Particular embodiments of the present invention can achieve the following advantages.
1. By adopting different risk metric value functions for different scenes, the accuracy of evaluation can be ensured.
2. The decision method for the dynamic behavior of the automatic driving vehicle in the internet environment combines the scene layering decision idea, fuses the safety guarantee function into the decision stage, can ensure the safety of decision in each step, considers the humanity of the automatic driving vehicle, introduces the soft constraints of an efficient soft constraint function, a comfortable soft constraint function and a traffic flow soft constraint function, and can further improve the intelligent level of the intelligent vehicle.
4. The invention makes real-time decision and driving decision by the driving control subsystem based on the information provided by the perception subsystem. Each algorithm is designed to be able to maneuver the vehicle under specific traffic conditions (e.g., road tracking, overtaking, intersection, etc.). Algorithms can be used to assist the driver (especially the driver of lower-ranked autonomous vehicles) in a number of ways, for example to warn the driver of an impending collision or to apply brakes or steering autonomously in critical traffic situations, or directly to the autonomous vehicle to make decision-making plans for vehicle behaviour.
Finally, it should be pointed out that: the above examples are only for illustrating the technical solutions of the present invention, and are not limited thereto. Those of ordinary skill in the art will understand that: modifications can be made to the technical solutions described in the foregoing embodiments, or some technical features may be equivalently replaced; such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (11)
1. A method for deciding the dynamic behavior of an automatic driving vehicle in an internet environment is characterized by comprising the following steps:
step S1, the surrounding road users acquire surrounding environment information under the V2X networking environment; taking the center of mass of the self-vehicle as the center, carrying out area division by using different radiuses, and estimating a risk area;
step S2, based on the surrounding environment information of the surrounding road users and the estimated risk area, making a first-stage action decision to determine a feasible action set which can be taken to ensure the driving safety of the vehicle;
step S3, performing a second stage action decision: and considering a non-safety constraint condition, optimally selecting a finally executed action from the feasible action set, and making a driving behavior decision.
2. The automated driving vehicle dynamic behavior decision method of claim 1, characterized in that in step S1, the risk area is estimated in the following way:
defining the safe braking distance L by taking the center of mass of the bicycle as the centerriskThe circular area of radius is the risk area,
in the formula, viIs the current speed of the vehicle, amaxThe maximum value of the acceleration of the bicycle is L, and the length of the bicycle is L; such asIf a certain surrounding road user is located in the risk area, the surrounding road user is defined as a risk road user;
defining the safety early warning distance L with the center of mass of the bicycle as the centerp-riskThe annular region after the risk region is removed for the circular region of radius is the potential risk region,
wherein, adecThe maximum value of the deceleration of the vehicle, if a certain surrounding road user is located in the potential risk area, the surrounding road user is defined as a potential risk road user;
defining the safety early warning distance L with the center of mass of the bicycle as the centerp-riskThe area outside the circular area of the radius is a safe area, and if a certain surrounding road user is located outside the potential risk area or outside the communication range of the own vehicle, the surrounding road user is defined as a safe road user.
3. The automated driving vehicle dynamic behavior decision method of claim 2, wherein step S2 comprises the steps of:
step S21, calculating a risk degree C for the risk area and the surrounding road users within the potential risk area, wherein the calculation is performed for the surrounding road users within the risk area first, then for the surrounding road users within the potential risk area,
the risk degree C characterizes the probability of a conflict between the current state of the autonomous vehicle and the state of the users of the surrounding roads,
t is the estimated time to collision, and if there are two or more surrounding road users in the risk area and the potential risk area, t is two or more estimates of the expected collision with each surrounding road userMinimum calculated predicted time to collision, t being greater than t if there are no surrounding road users in the risk zone and the potential risk zonecThe setting value of (a) is set,
tcthe critical time for avoiding collisions, for a set constant,
when C is 0, judging that no traffic conflict exists in the state, and determining a risk metric value friskGo to zero, go to step S23;
when C is 1, it is determined that there is a potential traffic collision in this state, go to step S22,
step S22, calculating a risk metric frisk，
Step S23, according to the risk metric friskAnd selecting the action and determining a feasible action set.
4. The automated driving vehicle dynamic behavior decision method of claim 3, wherein the estimated predicted time to collision t is calculated by the following equation,
t＝min{TTC,PET,TTB}
wherein,
Xiis the position of the vehicle at the self-parking position,
Xjis the position of the following surrounding road user,
viin order to obtain the current speed of the vehicle,
vjthe current speed of the other vehicle is the current speed,
Liin order to obtain the length of the bicycle,
PET is the time t for the vehicle to enter the conflict pointiTo the time t when another surrounding road user reaches the conflict pointjThe difference between the times of the two phases,
PET＝t＝|ti-tj|
TTB is used for evaluating the forward area, is suitable for the scene that the own vehicle is behind and the other vehicles are in front,
Xiis the position of the vehicle at the self-parking position,
Xjis the location of the other vehicle followed by,
viin order to obtain the current speed of the vehicle,
Liis the length of the bicycle.
5. The automated driving vehicle dynamic behavior decision method of claim 3, wherein the estimated predicted time to collision t is calculated in the following manner,
when a scene can be distinguished, calculating only an estimated predicted time-to-collision t for that scene, wherein for a straight-behind scene, t is TTC; for an intersection scene, t ═ PET; for the scene of collision of the self vehicle in the rear-front direction, t is TTB;
when the scene is complex, t is min { TTC, PET, TTB },
wherein,
Xiis the position of the vehicle at the self-parking position,
Xjis the position of the following surrounding road user,
viin order to obtain the current speed of the vehicle,
vjthe current speed of the other vehicle is the current speed,
Liin order to obtain the length of the bicycle,
PET is the time t for the vehicle to enter the conflict pointiTo the time t when another surrounding road user reaches the conflict pointjThe difference between the times of the two phases,
PET＝t＝|ti-tj|
TTB is used for evaluating the forward area, is suitable for the scene that the own vehicle is behind and the other vehicles are in front,
Xiis the position of the vehicle at the self-parking position,
Xjis the location of the other vehicle followed by,
viin order to obtain the current speed of the vehicle,
Liis the length of the bicycle.
6. The automated driving vehicle dynamic behavior decision method of claim 3, characterized in that in step S2, the risk metric value f is calculated by the following equationrisk，
Or
7. The automated driving vehicle dynamic behavior decision method according to claims 1-6, characterized in that in the second stage behavior decision of step S3, no decision attributes affecting safety are included, but rather an efficient soft constraint function f is consideredeComfort soft constraint function fcAnd a traffic flow soft constraint function ftAnd carrying out optimal decision.
8. The method of real-time trajectory planning for an autonomous vehicle of claim 7 wherein the high efficiency soft constraint function feIs defined as:
wherein, t0Is the initial departure time of the vehicle, tfV (t) is the speed of the vehicle at the time when the vehicle reaches the destination.
9. The method of real-time trajectory planning for an autonomous vehicle of claim 7 wherein comfortSoft constraint function fcIs defined as:
wherein a is the acceleration of the bicycle, alatFor lateral acceleration, alonIs the longitudinal acceleration.
10. The method of real-time trajectory planning for autonomous vehicles of claim 7 wherein the traffic flow soft constraint function ftIs defined as:
minft＝α(vave-vder)2+β(dave-dder)2
wherein,
vaveto determine the average speed level of the surrounding traffic flow before decision,
vderto determine the desired average speed level of the surrounding traffic flow,
daveto determine the average inter-vehicle distance of the surrounding traffic flow before decision,
dderto determine the desired average inter-vehicle distance for the surrounding traffic flow,
alpha and beta are weight coefficients which are both larger than 0 and smaller than 1.
11. The method for planning a real-time trajectory of an autonomous vehicle as claimed in claim 7, wherein in the second-stage action decision of step S3, the cost function J is defined as follows:
w1、w2、w3are weight coefficients, are all greater than 0 and less than 1, and w1+w2+w3＝1，
Wherein f ise0、fc0、ft0Respectively representing the safety, high efficiency, comfort and traffic flow functions after the self vehicle is supposed to continuously execute according to the state before decision.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910774762.XA CN110488802B (en) | 2019-08-21 | 2019-08-21 | Decision-making method for dynamic behaviors of automatic driving vehicle in internet environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910774762.XA CN110488802B (en) | 2019-08-21 | 2019-08-21 | Decision-making method for dynamic behaviors of automatic driving vehicle in internet environment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110488802A true CN110488802A (en) | 2019-11-22 |
| CN110488802B CN110488802B (en) | 2020-05-12 |
Family
ID=68552636
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910774762.XA Active CN110488802B (en) | 2019-08-21 | 2019-08-21 | Decision-making method for dynamic behaviors of automatic driving vehicle in internet environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110488802B (en) |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111615053A (en) * | 2020-05-22 | 2020-09-01 | 英华达（上海）科技有限公司 | Internet of vehicles system and method for dynamically marking risk areas |
| CN111775952A (en) * | 2019-11-26 | 2020-10-16 | 北京京东乾石科技有限公司 | Automatic driving method, device and storage medium |
| CN112017438A (en) * | 2020-10-16 | 2020-12-01 | 宁波均联智行科技有限公司 | Driving decision generation method and system |
| CN112150046A (en) * | 2020-11-06 | 2020-12-29 | 上海交通大学 | Road intersection safety risk index calculation method |
| CN112216146A (en) * | 2020-09-16 | 2021-01-12 | 浙江吉利控股集团有限公司 | Vehicle risk early warning method, device and equipment and vehicle |
| CN112256033A (en) * | 2020-10-30 | 2021-01-22 | 安徽江淮汽车集团股份有限公司 | Driving state recognition method, device, storage medium and device |
| CN112550284A (en) * | 2020-11-16 | 2021-03-26 | 东南大学 | Self-adaptive cruise driving risk assessment method under vehicle-road coupling effect |
| CN112885113A (en) * | 2019-11-29 | 2021-06-01 | 阿尔斯通运输科技公司 | Driving assistance method for public transport vehicle |
| CN112947390A (en) * | 2021-04-02 | 2021-06-11 | 清华大学 | Intelligent networking automobile safety control method and system based on environmental risk assessment |
| CN113022488A (en) * | 2019-12-25 | 2021-06-25 | 北京宝沃汽车股份有限公司 | Vehicle and control method and device thereof |
| CN113129624A (en) * | 2019-12-30 | 2021-07-16 | 百度（美国）有限责任公司 | Fastest lane determining algorithm under traffic jam condition |
| CN113177663A (en) * | 2021-05-20 | 2021-07-27 | 启迪云控(上海)汽车科技有限公司 | Method and system for processing intelligent network connection application scene |
| CN113433936A (en) * | 2021-06-02 | 2021-09-24 | 北京迈格威科技有限公司 | Mobile equipment obstacle-avoiding method and device, mobile equipment and storage medium |
| CN113548049A (en) * | 2021-07-27 | 2021-10-26 | 武汉理工大学 | Intelligent vehicle driving behavior decision method and system based on finite-state machine |
| CN113682299A (en) * | 2020-05-19 | 2021-11-23 | 广州汽车集团股份有限公司 | Vehicle forward collision early warning method and device |
| CN113954824A (en) * | 2020-07-02 | 2022-01-21 | 现代摩比斯株式会社 | Automatic driving system, method and readable medium for preventing collision with cut-in vehicle |
| CN114056332A (en) * | 2022-01-14 | 2022-02-18 | 清华大学 | Intelligent automobile following decision and control method based on cognitive risk balance |
| CN115331436A (en) * | 2022-08-04 | 2022-11-11 | 西安理工大学 | Road section automatic driving automobile movement planning method considering pedestrian safety |
| CN115620536A (en) * | 2022-10-18 | 2023-01-17 | 北京航空航天大学 | Method for improving crossing traffic efficiency based on danger degree in automatic driving environment |
| CN116978257A (en) * | 2023-08-11 | 2023-10-31 | 山东大学 | Highway confluence region conflict early warning method and system based on track prediction |
| US11878685B2 (en) | 2020-12-22 | 2024-01-23 | Beijing Baidu Netcom Science Technology Co., Ltd. | Vehicle control method and apparatus, electronic device and self-driving vehicle |
| CN117877272A (en) * | 2024-03-11 | 2024-04-12 | 中国市政工程华北设计研究总院有限公司 | Intersection safety evaluation method based on unmanned aerial vehicle detection |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103754224A (en) * | 2014-01-24 | 2014-04-30 | 清华大学 | Vehicle multi-target coordinating lane changing assisting adaptive cruise control method |
| CN105291965A (en) * | 2015-11-04 | 2016-02-03 | 石海水 | Automobile anti-collision method and device |
| CN109375632A (en) * | 2018-12-17 | 2019-02-22 | 清华大学 | Automatic driving vehicle real-time track planing method |
| CN109709956A (en) * | 2018-12-26 | 2019-05-03 | 同济大学 | A kind of automatic driving vehicle speed control multiple-objection optimization with algorithm of speeding |
| CN109910879A (en) * | 2019-04-03 | 2019-06-21 | 大连理工大学 | A kind of vehicle safety anticollision control method of combination safe distance and collision time |
| CN109941211A (en) * | 2019-03-22 | 2019-06-28 | 清华大学 | A kind of Intelligent Vehicle Driving System structure common type framework and construction method |
| CN110020504A (en) * | 2019-04-23 | 2019-07-16 | 吉林大学 | Unmanned vehicle running environment complexity quantization method based on rear intrusion and collision time |
| CN110136254A (en) * | 2019-06-13 | 2019-08-16 | 吉林大学 | Driving assistance information display methods based on dynamic probability driving map |
-
2019
- 2019-08-21 CN CN201910774762.XA patent/CN110488802B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103754224A (en) * | 2014-01-24 | 2014-04-30 | 清华大学 | Vehicle multi-target coordinating lane changing assisting adaptive cruise control method |
| CN105291965A (en) * | 2015-11-04 | 2016-02-03 | 石海水 | Automobile anti-collision method and device |
| CN109375632A (en) * | 2018-12-17 | 2019-02-22 | 清华大学 | Automatic driving vehicle real-time track planing method |
| CN109709956A (en) * | 2018-12-26 | 2019-05-03 | 同济大学 | A kind of automatic driving vehicle speed control multiple-objection optimization with algorithm of speeding |
| CN109941211A (en) * | 2019-03-22 | 2019-06-28 | 清华大学 | A kind of Intelligent Vehicle Driving System structure common type framework and construction method |
| CN109910879A (en) * | 2019-04-03 | 2019-06-21 | 大连理工大学 | A kind of vehicle safety anticollision control method of combination safe distance and collision time |
| CN110020504A (en) * | 2019-04-23 | 2019-07-16 | 吉林大学 | Unmanned vehicle running environment complexity quantization method based on rear intrusion and collision time |
| CN110136254A (en) * | 2019-06-13 | 2019-08-16 | 吉林大学 | Driving assistance information display methods based on dynamic probability driving map |
Non-Patent Citations (2)
| Title |
|---|
| XUNJIA ZHENG 等: "Behavioral decision-making model of the intelligent vehicle based on driving risk assessment", 《IEEE》 * |
| 宋威龙: "城区动态环境下智能车辆行为决策研究", 《中国博士学位论文全文数据库》 * |
Cited By (35)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111775952A (en) * | 2019-11-26 | 2020-10-16 | 北京京东乾石科技有限公司 | Automatic driving method, device and storage medium |
| CN112885113A (en) * | 2019-11-29 | 2021-06-01 | 阿尔斯通运输科技公司 | Driving assistance method for public transport vehicle |
| CN113022488A (en) * | 2019-12-25 | 2021-06-25 | 北京宝沃汽车股份有限公司 | Vehicle and control method and device thereof |
| US11880201B2 (en) | 2019-12-30 | 2024-01-23 | Baidu Usa Llc | Fastest lane determination algorithm under traffic jam |
| CN113129624A (en) * | 2019-12-30 | 2021-07-16 | 百度（美国）有限责任公司 | Fastest lane determining algorithm under traffic jam condition |
| CN113682299A (en) * | 2020-05-19 | 2021-11-23 | 广州汽车集团股份有限公司 | Vehicle forward collision early warning method and device |
| CN113682299B (en) * | 2020-05-19 | 2024-09-27 | 广州汽车集团股份有限公司 | Vehicle forward collision early warning method and device |
| CN111615053A (en) * | 2020-05-22 | 2020-09-01 | 英华达（上海）科技有限公司 | Internet of vehicles system and method for dynamically marking risk areas |
| CN113954824B (en) * | 2020-07-02 | 2024-03-01 | 现代摩比斯株式会社 | Automatic driving system, method and readable medium for preventing collision with cut-in vehicle |
| US11731616B2 (en) | 2020-07-02 | 2023-08-22 | Hyundai Mobis Co., Ltd. | Autonomous driving system for preventing collision of cut-in vehicle and autonomous driving method thereof |
| CN113954824A (en) * | 2020-07-02 | 2022-01-21 | 现代摩比斯株式会社 | Automatic driving system, method and readable medium for preventing collision with cut-in vehicle |
| CN112216146A (en) * | 2020-09-16 | 2021-01-12 | 浙江吉利控股集团有限公司 | Vehicle risk early warning method, device and equipment and vehicle |
| CN112017438B (en) * | 2020-10-16 | 2021-08-27 | 宁波均联智行科技股份有限公司 | Driving decision generation method and system |
| CN112017438A (en) * | 2020-10-16 | 2020-12-01 | 宁波均联智行科技有限公司 | Driving decision generation method and system |
| CN112256033B (en) * | 2020-10-30 | 2022-12-20 | 安徽江淮汽车集团股份有限公司 | Driving state recognition method, device, storage medium and device |
| CN112256033A (en) * | 2020-10-30 | 2021-01-22 | 安徽江淮汽车集团股份有限公司 | Driving state recognition method, device, storage medium and device |
| CN112150046A (en) * | 2020-11-06 | 2020-12-29 | 上海交通大学 | Road intersection safety risk index calculation method |
| CN112550284A (en) * | 2020-11-16 | 2021-03-26 | 东南大学 | Self-adaptive cruise driving risk assessment method under vehicle-road coupling effect |
| CN112550284B (en) * | 2020-11-16 | 2022-03-22 | 东南大学 | Self-adaptive cruise driving risk assessment method under vehicle-road coupling effect |
| US11878685B2 (en) | 2020-12-22 | 2024-01-23 | Beijing Baidu Netcom Science Technology Co., Ltd. | Vehicle control method and apparatus, electronic device and self-driving vehicle |
| CN112947390A (en) * | 2021-04-02 | 2021-06-11 | 清华大学 | Intelligent networking automobile safety control method and system based on environmental risk assessment |
| US11518409B2 (en) | 2021-04-02 | 2022-12-06 | Tsinghua University | Safety control method and system based on environmental risk assessment for intelligent connected vehicle |
| CN113177663B (en) * | 2021-05-20 | 2023-11-24 | 云控智行(上海)汽车科技有限公司 | Processing method and system of intelligent network application scene |
| CN113177663A (en) * | 2021-05-20 | 2021-07-27 | 启迪云控(上海)汽车科技有限公司 | Method and system for processing intelligent network connection application scene |
| CN113433936A (en) * | 2021-06-02 | 2021-09-24 | 北京迈格威科技有限公司 | Mobile equipment obstacle-avoiding method and device, mobile equipment and storage medium |
| CN113548049A (en) * | 2021-07-27 | 2021-10-26 | 武汉理工大学 | Intelligent vehicle driving behavior decision method and system based on finite-state machine |
| CN113548049B (en) * | 2021-07-27 | 2022-05-31 | 武汉理工大学 | Intelligent vehicle driving behavior decision method and system based on finite-state machine |
| CN114056332B (en) * | 2022-01-14 | 2022-04-12 | 清华大学 | Intelligent automobile following decision and control method based on cognitive risk balance |
| CN114056332A (en) * | 2022-01-14 | 2022-02-18 | 清华大学 | Intelligent automobile following decision and control method based on cognitive risk balance |
| CN115331436A (en) * | 2022-08-04 | 2022-11-11 | 西安理工大学 | Road section automatic driving automobile movement planning method considering pedestrian safety |
| CN115331436B (en) * | 2022-08-04 | 2024-04-30 | 西安理工大学 | Road section automatic driving automobile motion planning method considering pedestrian safety |
| CN115620536A (en) * | 2022-10-18 | 2023-01-17 | 北京航空航天大学 | Method for improving crossing traffic efficiency based on danger degree in automatic driving environment |
| CN116978257A (en) * | 2023-08-11 | 2023-10-31 | 山东大学 | Highway confluence region conflict early warning method and system based on track prediction |
| CN116978257B (en) * | 2023-08-11 | 2024-05-24 | 山东大学 | Highway confluence region conflict early warning method and system based on track prediction |
| CN117877272A (en) * | 2024-03-11 | 2024-04-12 | 中国市政工程华北设计研究总院有限公司 | Intersection safety evaluation method based on unmanned aerial vehicle detection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110488802B (en) | 2020-05-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110488802B (en) | Decision-making method for dynamic behaviors of automatic driving vehicle in internet environment | |
| EP3678911B1 (en) | Pedestrian behavior predictions for autonomous vehicles | |
| EP3828049B1 (en) | Machine readable storage medium and computing device | |
| JP6800899B2 (en) | Risk-based driver assistance for approaching intersections with limited visibility | |
| CN114222691A (en) | Cost-based path determination | |
| JP6308233B2 (en) | Vehicle control apparatus and vehicle control method | |
| US11256260B2 (en) | Generating trajectories for autonomous vehicles | |
| EP3854646A2 (en) | Systems and methods for navigating with safe distances | |
| US11814040B2 (en) | System and method for avoiding a collision course | |
| JP2023510136A (en) | Geolocation models for perception, prediction or planning | |
| JP6468261B2 (en) | Automated driving system | |
| US20230286536A1 (en) | Systems and methods for evaluating domain-specific navigation system capabilities | |
| JP2023533225A (en) | Methods and systems for dynamically curating autonomous vehicle policies | |
| WO2018132607A2 (en) | Navigation based on vehicle activity | |
| CN116249643A (en) | Method and system for predicting actions of an object by an autonomous vehicle to determine a viable path through a conflict area | |
| CN111295628A (en) | Semantic object clustering for autonomous vehicle decision making | |
| US20220314968A1 (en) | Electronic control device | |
| WO2023194801A1 (en) | Steering limiters for vehicle navigation | |
| JP2023533243A (en) | Vehicle independent trajectory verification system | |
| US20220332313A1 (en) | Method for avoiding a collision in road traffic on the basis of adaptively setting potentially occupied areas | |
| CN114132311A (en) | Method and module for screening dangerous targets for automatic emergency braking of vehicle | |
| JP2021160531A (en) | Vehicle control device, vehicle control method, and program | |
| Zhang et al. | Left-turn risk assessment of autonomous vehicles at unsignalized intersections | |
| EP4372715A1 (en) | Vehicle collision threat assessment | |
| CN114217601B (en) | Hybrid decision method and system for self-driving |
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 |