A lane-changing decision control method for autonomous vehicles considering uncertainty

By constructing observation and belief space models in autonomous vehicles, combining POMDP models and reward functions, decision-making problems under the influence of uncertainty are solved, more stable and reliable autonomous driving decisions are achieved, and the safety and comfort of the vehicle are improved. sex.

CN115257746BActive Publication Date: 2024-11-26TONGJI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210863061.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-21
Publication Date
2024-11-26
Estimated Expiration
2042-07-21

AI Technical Summary

Technical Problem

The existing decision-making methods of autonomous vehicles have little consideration of uncertainty caused by dynamic changes in the environment and bicycle state in the future, which makes it difficult to ensure the rationality and continuity of decision-making results, affecting the safety and comfort of the vehicle.

Method used

By constructing state space and action space, establishing observation space models and belief space models of bicycles and other vehicles, combining the set reward function, the set of decision state points of bicycles is solved based on the POMDP model and decoupling them into horizontal space The decision set and the vertical time decision set introduce road boundaries, vehicle speed and obstacle distance constraints, divide the travelable area and reference path, and output the optimal trajectory.

Benefits of technology

It improves the reliability and stability of decision-making results, enhances the comfort and safety of driving vehicles, and is suitable for typical driving conditions such as following cars, changing lanes and overtaking independently.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115257746B_ABST
    Figure CN115257746B_ABST
Patent Text Reader

Abstract

The present invention relates to a lane-changing decision control method for an autonomous driving vehicle considering uncertainty, including: constructing a state space and an action space, establishing a state transfer equation; establishing an observation space model and a belief space model of the self-vehicle and other-vehicle, and an uncertainty model of the predicted trajectory of the other-vehicle; setting a reward function, combining a POMDP model, solving the decision state point set of the self-vehicle, and decoupling it into a lateral space decision set and a longitudinal time decision set; respectively determining the lateral and longitudinal drivable boundaries, introducing road boundary constraints, vehicle speed constraints and obstacle distance constraints, dividing the lateral drivable area and reference path, longitudinal drivable area and reference speed curve; according to the decision result, the vehicle planning module outputs the optimal trajectory of the corresponding vehicle, so that the vehicle drives according to the optimal trajectory. Compared with the prior art, the present invention fully considers the uncertainty of the predicted trajectory of other-vehicles, and the self-vehicle state point set determined by the decision is more stable and reliable, which can effectively enhance the comfort and safety of vehicle driving.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving vehicle control, and in particular to a lane changing decision control method for an autonomous driving vehicle taking uncertainty into consideration. Background Art

[0002] Decision planning is one of the key parts of autonomous driving. It first integrates multi-sensor information, then makes task decisions based on driving needs, and then plans multiple optional safe paths between two points under the premise of avoiding possible obstacles and through some specific constraints, and selects the best path among these paths as the vehicle's driving trajectory.

[0003] At present, the decision-making module of the vehicle comprehensively collects the required information and makes reasonable driving decisions after comprehensive analysis, while the planning module of the vehicle outputs a safe and comfortable vehicle trajectory based on the decision results. In existing research, the more representative decision-making methods include finite state machines, behavior trees, Markov processes, etc. However, most decision-making methods give little consideration to the uncertainty problems caused by the dynamic changes of the environment and the vehicle state in the future, and most decision-making methods simply output behavioral instructions, which cannot guarantee the rationality of the decision results. There may even be discontinuities in the decision results in time, which brings difficult solutions to the trajectory planning module, which can easily lead to poor vehicle safety and comfort and poor user experience. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide an autonomous driving vehicle lane changing decision control method taking into account uncertainty. By considering the uncertainty of the predicted trajectory of other vehicles, the reliability of the decision results is improved to enhance the comfort and safety of driving the vehicle.

[0005] The purpose of the present invention can be achieved by the following technical solution: A lane-changing decision control method for an autonomous driving vehicle considering uncertainty, comprising the following steps:

[0006] S1. Based on the perceived state of the vehicle and the predicted state of other vehicles, combined with the vehicle dynamics model, the state space and action space are constructed, and the state transfer equation is established;

[0007] S2, respectively establish the observation space model and belief space model of the own vehicle, the observation space model and belief space model of the other vehicle, and the uncertainty model of the predicted trajectory of the other vehicle;

[0008] S3, set the reward function, combine steps S1 and S2 to construct a POMDP (Partially Observable Markov Decision Process) model, and solve the decision state point set of the vehicle;

[0009] S4, decoupling the decision state point set of the vehicle into a lateral spatial decision set and a longitudinal temporal decision set;

[0010] S5. Determine the lateral drivable boundary, introduce road boundary constraints, vehicle speed constraints and obstacle distance constraints, and divide the lateral drivable area and reference path;

[0011] Determine the longitudinal drivable boundary, introduce road boundary constraints, vehicle speed constraints and obstacle distance constraints, and divide the longitudinal drivable area and reference speed curve;

[0012] S6. According to the lateral drivable area and reference path, longitudinal drivable area and reference speed curve divided in step S5, the vehicle planning module outputs the optimal trajectory of the corresponding vehicle so that the vehicle travels according to the optimal trajectory.

[0013] Furthermore, the step S1 specifically includes the following steps:

[0014] S11. Use a simplified vehicle kinematic model to simplify the vehicle's motion into the motion of a mass point in the Frenet coordinate system to construct the state space and action space of the ego vehicle;

[0015] S12. Use the state space and action space to establish the state transfer function of the vehicle and other vehicles.

[0016] Furthermore, the step S11 specifically includes the following steps:

[0017] S111, collecting the perceived vehicle position, longitudinal speed, longitudinal acceleration, and lateral speed information of the ego vehicle to form a state space of the ego vehicle;

[0018] Collect the predicted heading angle, vehicle position, longitudinal speed, longitudinal acceleration, and lateral speed information of the other vehicle, and combine it with the length and width of the vehicle body to form the state space of the other vehicle;

[0019] S112: Setting a discrete lateral acceleration sequence and a discrete longitudinal acceleration sequence as an action space.

[0020] Furthermore, the specific process of step S12 is: assuming that the evolution processes of the states of the own vehicle and the other vehicle are independent of each other, the state transfer equation of the own vehicle is obtained according to the simplified vehicle motion model, and the state transfer equation of the other vehicle is expressed by the probabilities of different states at the next moment.

[0021] Furthermore, the observation function of the observation space model in step S2 is divided into the own vehicle and the other vehicle, wherein the own vehicle observation function is represented by a Boolean value, which is set to 1 if the next state exists, otherwise it is set to 0;

[0022] The observation function of the other car conforms to the Gaussian distribution;

[0023] The uncertainty model of the predicted trajectory of the other vehicle in step S2 is constructed using multivariate Gaussian distribution.

[0024] Furthermore, the step S3 specifically includes the following steps:

[0025] S31, setting the reward function including the safety reward function R safe , comfort reward function R comfort And the efficiency reward function R efficiency , where the safety reward function R safe Including the collision reward function R colli And the distance reward function R dis ;

[0026] Comfort reward function R comfort Including speed-related comfort reward function R speed , lateral speed action penalty function R vlat and the continuity index R continuity ;

[0027] Efficiency reward function R efficiency Including the target task reward function R lane and the target speed function R v_tar ;

[0028] S32. Based on the constructed state space, action space, belief space and reward function, the deterministic sparse partially observable tree (DESPOT) method is used to solve the POMDP model and obtain the decision state point set of the vehicle, where the content of the decision state point set includes the position information of the vehicle and the speed and acceleration information at different times.

[0029] Furthermore, the step S31 specifically designs a collision reward function R according to the probability of collision colli , that is, when the collision probability exceeds the set threshold, a penalty is given;

[0030] The step S31 specifically designs a distance reward function R according to the collision time (TTC) model dis , and set the upper and lower limits of the maximum safe distance;

[0031] In step S31, the speed-related comfort reward function R speed The influencing factors include the current longitudinal velocity vlon cur , longitudinal acceleration alon cur , lateral speed vlat cur And the longitudinal velocity vlon at the next moment next , longitudinal acceleration alon next , lateral speed vlat next ;

[0032] In step S31, the lateral speed action penalty function R vlat It is only related to the lateral speed;

[0033] In step S31, the continuity index R continuity It is positively correlated with the position change between the two decisions. The greater the position change between the two decisions, the greater the R continuity The bigger.

[0034] Furthermore, in step S4, the lateral spatial decision set includes the position information of the vehicle; and the longitudinal temporal decision set includes the speed and acceleration information of the vehicle.

[0035] Furthermore, when the lateral drivable boundary is selected in step S5, the distance from the node of the lateral drivable boundary to the obstacle should be greater than the safety distance, and the lateral drivable boundary value should not exceed the original boundary of the structured road, and a safety threshold distance should be maintained;

[0036] The process of establishing the horizontal reference path is specifically as follows:

[0037] First, a lateral optimization problem is constructed, and a lateral cost function is designed to evaluate each node. The point where the lateral cost function is the minimum is taken as the optimal solution. The constraint of the lateral optimization problem is that the heading angle is between the minimum and maximum wheel turning angles. The objective function of the lateral optimization problem is the lateral cost function. The lateral cost function C node-h is the first distance cost C d , the first safety cost C o and continuity costs C c The weighted sum of

[0038] By calculating the distance from the node to the target state point as the first distance cost; calculating the distance from the node to the obstacle as the first safety cost; calculating the position change rate between the front and rear nodes as the continuity cost; combining the corresponding weight coefficients, the lateral cost function is calculated;

[0039] Finally, the value at the minimum of the lateral cost function is taken, and the lines are connected to form the lateral expected reference path.

[0040] Furthermore, the principle for determining the longitudinal drivable boundary in step S5 is: the longitudinal drivable boundary coincides with the position coordinates of the obstacle vehicle;

[0041] The constraint conditions of the longitudinal drivable boundary are: the st curve of the longitudinal drivable boundary does not exceed the st curves representing the highest and lowest average vehicle speeds;

[0042] The process of establishing the longitudinal reference speed curve is specifically as follows:

[0043] First, a vertical optimization problem is constructed, and a vertical cost function is designed to evaluate each node. The point where the vertical cost function is the minimum is taken as the optimal solution. node-z is the second distance cost C d-ref , the second safety cost C o-ref and speed change cost C v The weighted sum of

[0044] The distance from the node to the state point is calculated as the second distance cost; the square of the difference between the distance from the node to the obstacle on the s-axis and the distance threshold to the obstacle is calculated as the second safety cost; the rate of change of the reference speed is calculated as the speed change cost; combined with the corresponding weight coefficient, the longitudinal cost function is calculated;

[0045] Finally, the value at the minimum of the longitudinal cost function is taken, and the lines are connected to form the longitudinal expected reference speed curve.

[0046] Compared with the prior art, the present invention fully considers the uncertainty of the predicted trajectory of other vehicles, establishes the observation space model and belief space model of the self-vehicle and other vehicles, establishes the uncertainty model of the predicted trajectory of other vehicles, and then combines the set reward function to solve the decision state point set of the self-vehicle based on the POMDP model. This makes the state point set of the self-vehicle determined by the decision more stable and reliable, which is conducive to enhancing the comfort and safety of driving the vehicle, and can be well applied to unmanned driving under typical driving conditions such as following the vehicle, changing lanes and autonomous overtaking.

[0047] After solving the decision state point set of the vehicle, the present invention decouples it into a lateral spatial decision set and a longitudinal temporal decision set. By introducing road boundary constraints, vehicle speed constraints and obstacle distance constraints, the lateral drivable area and the reference path, and the longitudinal drivable area and the reference speed curve are divided respectively, which can further ensure the reliability of the decision result. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0049] Figure 2 Schematic diagram of lane change decision process in the embodiment;

[0050] Figure 3 Schematic diagram of uncertainty of the predicted trajectory of another vehicle in the embodiment;

[0051] Figure 4 A schematic diagram of a lateral search to determine a lateral drivable boundary in an embodiment;

[0052] Figure 5 A diagram showing a lateral drivable area and a lateral expected reference path in the embodiment;

[0053] Figure 6 Graph showing the longitudinal drivable area and longitudinal expected reference speed in the embodiment. DETAILED DESCRIPTION

[0054] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] Example

[0056] like Figure 1 As shown, a lane-changing decision control method for an autonomous driving vehicle considering uncertainty includes the following steps:

[0057] S1. Based on the perceived state of the vehicle and the predicted state of other vehicles, combined with the vehicle dynamics model, the state space and action space are constructed, and the state transfer equation is established;

[0058] S2, respectively establish the observation space model and belief space model of the own vehicle, the observation space model and belief space model of the other vehicle, and the uncertainty model of the predicted trajectory of the other vehicle;

[0059] S3, set the reward function, combine steps S1 and S2 to construct the POMDP model, and solve the decision state point set of the vehicle;

[0060] S4, decoupling the decision state point set of the vehicle into a lateral spatial decision set and a longitudinal temporal decision set;

[0061] S5. Determine the lateral drivable boundary, introduce road boundary constraints, vehicle speed constraints and obstacle distance constraints, and divide the lateral drivable area and reference path;

[0062] Determine the longitudinal drivable boundary, introduce road boundary constraints, vehicle speed constraints and obstacle distance constraints, and divide the longitudinal drivable area and reference speed curve;

[0063] S6. According to the lateral drivable area and reference path, longitudinal drivable area and reference speed curve divided in step S5, the vehicle planning module outputs the optimal trajectory of the corresponding vehicle so that the vehicle travels according to the optimal trajectory.

[0064] This embodiment applies the above technical solution, and its lane change decision process is as follows: Figure 2As shown, including:

[0065] 1. Based on the perceived state of the vehicle and the predicted state of other vehicles, the state space and action space are constructed according to the simplified vehicle dynamics model, and the state transfer equation is established;

[0066] Specifically:

[0067] 11) Using a simplified vehicle kinematic model, the vehicle's motion is simplified to the motion of a particle in the Frenet coordinate system, and the state space and action space of the vehicle are constructed;

[0068] 111) The state space expression is:

[0069] S=[state ego ,state 1 ,state 2 ,...,state n ]

[0070]

[0071]

[0072] Among them, state ego is the state space of the vehicle, time is the time mark, (s ego ,l ego ) is the position of the vehicle, vlon ego is the longitudinal velocity of the vehicle, acc ego is the longitudinal acceleration of the vehicle, vlat ego is the lateral speed information of the vehicle;

[0073] state n is the state space of the other car n, (length, width) is the length and width of the other car n, θ is the predicted heading angle of the other car, (s, l) is the position of the other car, v n The speed information of his car;

[0074] 112) Set a discrete lateral acceleration sequence and a longitudinal acceleration sequence as the action space:

[0075] A=[Acc lon ,Vel lat ]

[0076] Vel vlat ={-2.0,-1.5,-1.0,-0.5,0.0,0.5,1.0,1.5,2.0}

[0077] Acc lon={-3.0,-2.5,-2.0,-1.5,-1.0,-0.5,0.0,0.5,1.0,1.5,2.0}

[0078] Among them, Vel vlat is the lateral acceleration sequence; Acc lon is the longitudinal acceleration sequence;

[0079] 12) Using the state space and action space, establish the state transfer function of the self-vehicle and other vehicles;

[0080] The specific establishment process is as follows:

[0081] 121) Assuming that the evolution of the states of the ego vehicle and the other vehicle are independent of each other, the state transition equation of the ego vehicle is obtained based on the simplified vehicle motion model, and the state transition equation of the other vehicle is expressed by the probability of different states at the next moment;

[0082] The calculation formula of the vehicle state transfer equation is:

[0083]

[0084] The calculation formula of the state transfer equation of other vehicles is:

[0085]

[0086] 2. Establish the observation space model and belief space model of the self-vehicle and other vehicles, and establish the uncertainty model of the predicted trajectory of other vehicles. The observation space contains the information that can be observed by the intelligent vehicle environment perception and positioning system, including the position, speed, and heading angle information of the self-vehicle and the position, speed, and heading angle information of other vehicles;

[0087] The specific steps are as follows:

[0088] 21) The state of the self-vehicle is completely observable, and the observation function of the other vehicle includes coordinates and heading angles, which conform to Gaussian distribution;

[0089] Establish the observation function of the observation space. The ego vehicle observation function is represented by a Boolean value. If the next state exists, it is set to 1, otherwise it is 0:

[0090] The observation function of the ego vehicle is:

[0091]

[0092] The observation function of other cars conforms to the Gaussian distribution, and the calculation formula is:

[0093]

[0094] where μ s,l,θ represents the observed value, σ s,l,θ represents variance;

[0095] The process of calculating the pose uncertainty through the observation model is:

[0096] 22) Use multivariate Gaussian distribution to construct the uncertainty model of the other vehicle and the vehicle’s posture;

[0097] The uncertainty calculation formula of the other vehicle's own vehicle posture is:

[0098]

[0099] Among them, Φ is the state transfer matrix, M represents the noise in the state transfer process, Z is the covariance of the noise, Σ is the covariance matrix of the system, and X is the state quantity, which is expressed as:

[0100]

[0101] Through uncertainty modeling, the Gaussian distribution covariance of each state quantity at the predicted trajectory point of the other vehicle is obtained as follows: Figure 3 As shown;

[0102] 3. Set the reward function and solve the decision state point set of the vehicle based on the POMDP model;

[0103] 31) Set the content of the reward function: safety index R safe , comfort index R comfort and efficiency index R efficiency ;

[0104] 32) Design a reward function for safety indicators: Consider the collision reward function R colli And the distance reward function R dis , the specific calculation formula is:

[0105] R safe =R colli +R dis

[0106] 33) Design a collision reward function R based on the probability of collision colli , when the collision probability exceeds the threshold, a penalty is given. Specifically, the collision reward function is solved as follows:

[0107]

[0108] Among them, w colli =100 is the weight value, P safe is the safety threshold;

[0109] Design the reward function R of the front vehicle distance based on the collision time (TTC) model dis , set the upper and lower limits of the maximum safety distance. Specifically, the reward function for the distance from the front vehicle is calculated as follows:

[0110]

[0111] Among them, d max d min are the maximum safety distance and minimum safety distance between the vehicle and other vehicles, respectively, and w dis =20 is the weight value of the distance reward;

[0112] 34) Design a comfort reward function: Consider the speed-related comfort reward function R speed , lateral speed action penalty function R vlat and the continuity index is R continuity , the comfort reward function is written as:

[0113] R comfort =R speed +R vlat +R continuity

[0114] Among them, the speed-related comfort reward function R speed The solution formula is:

[0115]

[0116] The current vertical speed is vlon cur , the current longitudinal acceleration is alon cur , the current lateral speed is vlat cur , the longitudinal velocity at the next moment is vlon next , the longitudinal acceleration at the next moment is alon next , the lateral velocity at the next moment is vlat next , w speed =15 is the weight coefficient of speed;

[0117] The calculation formula of the lateral speed action penalty function is:

[0118] R vlat =-vlat 2 × vlat

[0119] Among them, w vlat =12 is the weight coefficient of lateral speed;

[0120] The calculation formula of the continuity index of the result is:

[0121]

[0122] Among them, t represents the time of each step of the decision result; m and n represent the total number of steps of the previous and next decision results, respectively. last_loop,t ,llast_loop,t ) and (s cur_loop,t ,l cur_loop,t ) are the location information of the last and current decision results, respectively, w con =12 is the weight coefficient of the continuity index;

[0123] 35) Design efficiency reward function R efficiency :Consider the target task reward function R lane and the target speed function R v_tar , the efficiency reward function is written as:

[0124] R efficiency =R lane +R v_tar

[0125] Among them, the target task reward function is R lane =-|l lane -l ego |×w lane ;

[0126] The target speed function is R v_tar =-|v tar -v ego |×w v_tar ;

[0127] w v_tar =12 is the weight coefficient of target speed;

[0128] 36) Solve the decision action sequence and divide the decision into a horizontal space decision set and a vertical space decision set. Specifically, the POMDP model is solved by using a deterministic sparse observable tree method to obtain a set of ego vehicle decision state points including the ego vehicle position information and the speed and acceleration information at different times;

[0129] 4. Decouple the decision state point set into a lateral spatial decision set and a longitudinal temporal decision set, where the lateral spatial decision set includes the position information of the ego vehicle at each moment; the longitudinal temporal decision set includes the speed and acceleration information of the ego vehicle at each moment;

[0130] 5. Determine the lateral drivable boundary, introduce road boundary constraints, vehicle speed constraints and obstacle distance constraints, and divide the lateral drivable area and reference path;

[0131] When selecting the lateral drivable boundary, the distance from the boundary node to the obstacle should be greater than the safety distance. The specific expression is:

[0132]

[0133] Among them, d obsis the distance between the current node and the obstacle. When it is less than a certain safety distance, the node is considered unsafe. k is the lateral coordinate of the node, which should satisfy l min ≤l k ≤l max ;

[0134] At the same time, the drivable boundary value should not exceed the boundary value l of the original boundary of the structured road. min With l max , and maintain a safety threshold d safe Distance:

[0135]

[0136] The lateral drivable area obtained after calculation is as follows: Figure 4 As shown;

[0137] A lateral optimization model is established in the lateral drivable area to obtain a lateral reference path, specifically:

[0138] Construct a horizontal optimization problem, design a horizontal cost function to evaluate each node, and the point with the minimum horizontal cost function is the optimal point;

[0139] The constraint is a heading angle constraint:

[0140]

[0141] Among them, θ' min ,θ' max are the minimum and maximum values ​​of the vehicle's heading angle change, respectively;

[0142] The horizontal cost function is used as the objective function, and the expression is:

[0143] C node-h =w d C d +w o C o +w c C c

[0144] Among them, C d is the first distance cost of the node, C o is the first safety cost, C c is the continuity cost, and the corresponding weight functions are w d =0.35, w o =0.45, w c =0.2;

[0145] The first distance cost function C d The expression is:

[0146]

[0147] Among them, (s state ,l state ) is the position of the target state point, (s k ,l k ) is the location of the expansion node;

[0148] The first safety cost function C o The expression is:

[0149]

[0150] Among them, d obs Represents the distance from the node to the nearest obstacle, d max Represents the distance threshold to the obstacle;

[0151] The cost function C represents the continuity c The expression is:

[0152]

[0153] Among them, l i ,l i+1 ,l i+2 is the lateral position of the front and rear nodes;

[0154] Finally, the value at the minimum of the lateral cost function is connected to form the lateral expected reference path;

[0155] The lateral desired reference path formed in this embodiment is Figure 5 Hollow dots are used to represent;

[0156] 6. Determine the longitudinal drivable boundary, introduce road boundary constraints, vehicle speed constraints and obstacle distance constraints, and divide the longitudinal drivable area and reference speed curve;

[0157] The specific method for determining the longitudinal drivable boundary is as follows:

[0158] Ensure the boundary s in the s-axis direction border Satisfied:

[0159]

[0160] s border The constraints are: min_t ≤s border ≤s max_t , where s max_t and min_t are the upper and lower bounds of the average velocity, respectively;

[0161] The specific method for establishing the longitudinal reference speed curve is as follows:

[0162] Construct a vertical optimization problem, design a vertical cost function to evaluate each node, and the point where the vertical cost function is the minimum is the optimal point; the vertical cost function calculation formula is:

[0163] C node-z =w d-ref C d-ref +w o-ref C o-ref +w v C v

[0164] Among them, C d-ref is the second distance cost of the line from the node to the state point, C o-ref is the second security cost of the node, C v is the cost of speed change, and the corresponding weight coefficients are w d-ref =0.4, w o-ref =0.4, w v =0.2.

[0165] The second distance cost function is:

[0166] c d-ref =(s i -s ref ) 2

[0167] Among them, s ref is the position of the node on the line connecting the two state points;

[0168] The second safety cost function is:

[0169]

[0170] Among them, s obs Represents the distance from the node to the nearest obstacle along the s axis, s max Represents the distance threshold to the obstacle;

[0171] The cost function for speed change is:

[0172]

[0173] Finally, the value at the minimum of the longitudinal cost function is connected to form the longitudinal expected reference speed curve.

[0174] The longitudinal reference speed curve formed in this embodiment is as follows: Figure 6 shown.

[0175] In summary, this technical solution takes into account the behavioral interaction characteristics of the ego vehicle and traffic participants and the uncertainty of the predicted trajectory of other vehicles. The decision-making process is more stable, and the determined ego vehicle state point set is more reasonable and reliable, which is conducive to the subsequent vehicle planning module to solve accurately and quickly, thereby effectively enhancing the comfort and safety of vehicle driving.

Claims

1. A lane-changing decision control method for an autonomous driving vehicle considering uncertainty, characterized in that: The following steps are involved: S1. Based on the perceived state of the vehicle and the predicted state of other vehicles, combined with the vehicle dynamics model, the state space and action space are constructed, and the state transfer equation is established; S2, respectively establish the observation space model and belief space model of the own vehicle, the observation space model and belief space model of the other vehicle, and the uncertainty model of the predicted trajectory of the other vehicle; S3, set the reward function, combine steps S1 and S2 to construct the POMDP model, and solve the decision state point set of the vehicle; S4, decoupling the decision state point set of the vehicle into a lateral spatial decision set and a longitudinal temporal decision set; S5. Determine the lateral drivable boundary, introduce road boundary constraints, vehicle speed constraints and obstacle distance constraints, and divide the lateral drivable area and reference path; Determine the longitudinal drivable boundary, introduce road boundary constraints, vehicle speed constraints and obstacle distance constraints, and divide the longitudinal drivable area and reference speed curve; S6, according to the lateral drivable area and reference path, longitudinal drivable area and reference speed curve divided in step S5, the vehicle planning module outputs the optimal trajectory of the corresponding vehicle, so that the vehicle drives according to the optimal trajectory; The step S1 specifically includes the following steps: S11. Use a simplified vehicle kinematic model to simplify the vehicle's motion into the motion of a mass point in the Frenet coordinate system to construct the state space and action space of the ego vehicle; S12, using the state space and action space, establish a state transfer function for transferring the self-vehicle and other vehicles; The step S11 specifically includes the following steps: S111, collecting the perceived vehicle position, longitudinal speed, longitudinal acceleration, and lateral speed information of the ego vehicle to form a state space of the ego vehicle; Collect the predicted heading angle, vehicle position, longitudinal speed, longitudinal acceleration, and lateral speed information of the other vehicle, and combine it with the length and width of the vehicle body to form the state space of the other vehicle; S112, setting a discrete lateral acceleration sequence and a longitudinal acceleration sequence as an action space; The observation function of the observation space model in step S2 is divided into the own vehicle and other vehicles, wherein the own vehicle observation function is represented by a Boolean value, which is set to 1 if the next state exists, otherwise it is set to 0; The observation function of the other car conforms to the Gaussian distribution; The uncertainty model of the predicted trajectory of the other vehicle in step S2 is constructed using multivariate Gaussian distribution; The step S3 specifically comprises the following steps: S31, setting the reward function including the safety reward function R safe , comfort reward function R comfort And the efficiency reward function R efficiency , where the safety reward function R safe Including the collision reward function R colli And the distance reward function R dis ; Comfort reward function R comfort Including speed-related comfort reward function R speed , lateral speed action penalty function R vlat and the continuity index R continuity ; Efficiency reward function R efficiency Including the target task reward function R lane and the target speed function R v_tar ; S32. Based on the constructed state space, action space, belief space and reward function, the deterministic sparse observable tree method is used to solve the POMDP model and obtain the decision state point set of the vehicle, where the content of the decision state point set includes the position information of the vehicle and the speed and acceleration information at different times.

2. The lane changing decision control method for an autonomous driving vehicle considering uncertainty according to claim 1, characterized in that: The specific process of step S12 is: assuming that the evolution processes of the states of the own vehicle and the other vehicle are independent of each other, the state transfer equation of the own vehicle is obtained according to the simplified vehicle motion model, and the state transfer equation of the other vehicle is expressed by the probabilities of different states at the next moment.

3. The lane changing decision control method for an autonomous driving vehicle considering uncertainty according to claim 1, characterized in that: The step S31 specifically designs a collision reward function R according to the probability of collision. colli , that is, when the collision probability exceeds the set threshold, a penalty is given; The step S31 specifically designs a distance reward function R according to the collision time (TTC) model dis , and set the upper and lower limits of the maximum safe distance; In step S31, the speed-related comfort reward function R speed The influencing factors include the current longitudinal velocity vlon cur , longitudinal acceleration alon cur , lateral speed vlat cur And the longitudinal velocity vlon at the next moment next , longitudinal acceleration alon next , lateral speed vlat next ; In step S31, the lateral speed action penalty function R vlat Only related to lateral speed; In step S31, the continuity index R continuity It is positively correlated with the position change between the two decisions. The greater the position change between the two decisions, the greater the R continuity The bigger.

4. The lane changing decision control method for an autonomous driving vehicle considering uncertainty according to claim 1, characterized in that: In step S4, the lateral spatial decision set includes the position information of the vehicle; the longitudinal temporal decision set includes the speed and acceleration information of the vehicle.

5. The lane changing decision control method for an autonomous driving vehicle considering uncertainty according to claim 1, characterized in that: When selecting the lateral drivable boundary in step S5, the distance from the node of the lateral drivable boundary to the obstacle is greater than the safety distance, and the lateral drivable boundary value does not exceed the original boundary of the structured road and maintains a safety threshold distance; The specific process of establishing the reference path of the lateral drivable area is as follows: First, a lateral optimization problem is constructed, and a lateral cost function is designed to evaluate each node. The point where the lateral cost function is the minimum is taken as the optimal solution. The constraint of the lateral optimization problem is that the heading angle is between the minimum and maximum wheel turning angles. The objective function of the lateral optimization problem is the lateral cost function. The lateral cost function is the first distance cost C d , the first safety cost C o and continuity costs C c The weighted sum of By calculating the distance from the node to the target state point as the first distance cost; calculating the distance from the node to the obstacle as the first safety cost; calculating the position change rate between the front and rear nodes as the continuity cost; combining the corresponding weight coefficients, the lateral cost function is calculated; Finally, the value at the minimum of the lateral cost function is taken, and the lines are connected to form the lateral expected reference path.

6. The lane-changing decision control method for an autonomous driving vehicle considering uncertainty according to claim 5, characterized in that: The determination principle of the longitudinal drivable boundary in step S5 is: the longitudinal drivable boundary coincides with the position coordinates of the obstacle vehicle; The constraint conditions of the longitudinal drivable boundary are: the st curve of the longitudinal drivable boundary does not exceed the st curves representing the highest and lowest average vehicle speeds; The specific process of establishing the reference speed curve of the longitudinal drivable area is as follows: First, a longitudinal optimization problem is constructed, and a longitudinal cost function is designed to evaluate each node. The point where the longitudinal cost function is the minimum is taken as the optimal solution. The longitudinal cost function is the second distance cost C d-ref , the second safety cost C o-ref and speed change cost C v The weighted sum of The distance from the node to the state point is calculated as the second distance cost; the square of the difference between the distance from the node to the obstacle on the s-axis and the distance threshold to the obstacle is calculated as the second safety cost; the rate of change of the reference speed is calculated as the speed change cost; combined with the corresponding weight coefficient, the longitudinal cost function is calculated; Finally, the value at the minimum of the longitudinal cost function is taken, and the lines are connected to form the longitudinal expected reference speed curve.