A dynamic path planning method for autonomous driving vehicles

Abstract

The present invention discloses a method for dynamic path planning of an autonomous driving vehicle. The specific contents are as follows: 1. Establishing a driving environment model including road conditions, vehicle speed and position information; 2. Designing a hierarchical fuzzy reasoning system in combination with the driving environment model to determine the type of path that needs to be planned at the current position; 3. Constructing a vehicle driving path based on a B-spline curve to obtain pending parameters that need to be optimized; 4. Considering the two factors of distance and curvature, designing an optimization solution method based on a particle swarm to solve the optimal path; 5. Using a fuzzy fitting method, calculating the reference speed corresponding to the above optimal path. The path planning work performed by the present invention can complete the driving decision of the autonomous driving vehicle in a dynamic environment, adapt to different road conditions and vehicle speeds, and improve the lateral dynamic stability of the vehicle during driving.

Description

Technical Field

[0001] The present invention belongs to the field of path planning and relates to a dynamic path planning method for an autonomous vehicle. Background Art

[0002] Safety is always the top priority when driving a vehicle. Statistical studies show that 94% of traffic accidents are caused by driving failures, including distracted driving, fatigue driving, and emotional driving. Autonomous driving technology provides an effective technical solution to solve these problems. In particular, the research and application of advanced driver assistance systems (ADAS) have greatly reduced potential traffic accidents and improved driving safety. Therefore, the industry and academia have been actively involved in the development of autonomous vehicles, and these technologies have improved vehicle safety, alleviated traffic congestion, and optimized driving behavior. However, there are still many difficulties to overcome, mainly including perception, planning, and control.

[0003] Among them, path planning is one of the most critical and complex problems. However, current path planning algorithms do not consider the influence of road surface friction coefficient and vehicle speed on the vehicle. However, when the friction coefficient is low or the speed of the autonomous vehicle is high, it is easy to lose lateral dynamic stability, causing the vehicle to lose control and resulting in serious traffic accidents.

[0004] In view of this, the present invention provides a dynamic path planning method for an autonomous vehicle to adapt to different road conditions and vehicle speeds and improve the lateral dynamic stability during vehicle driving. Summary of the Invention

[0005] An embodiment of the present invention provides a dynamic path planning method for an autonomous vehicle, including:

[0006] Step S1, normalizing driving environment factors to establish a driving environment model;

[0007] Step S2, combining the driving environment model to construct a hierarchical fuzzy inference system, dynamically updating environment parameters, and calculating the path type to be planned at the current position and the starting position of path planning, where the path type includes executing a turn or going straight;

[0008] Step S3, constructing a vehicle driving path based on a B-spline curve according to vehicle dynamics characteristics to obtain undetermined parameters to be optimized;

[0009] Step S4, considering two factors of distance and curvature, designing an optimization method based on particle swarm to optimize and solve the undetermined parameters in the vehicle path optimization model to obtain an optimal path;

[0010] Step S5, using the method of fuzzy fitting to calculate the vehicle driving paths under different road conditions and the corresponding reference speeds.

[0011] Preferably, the driving environment factors at least include: road surface friction, target vehicle speed, and the speed difference between the target vehicle and the environmental vehicle.

[0012] Preferably, in the step S1, the driving environment model is obtained by normalizing the road surface friction, target vehicle speed, and the speed difference between the target vehicle and the environmental vehicle. Among them, the normalization method for the target vehicle speed is:

[0013]

[0014] In the formula, v min and v max are the minimum speed and maximum speed of the target vehicle, which are set to 0 km / h and 120 km / h, v' subj is the target vehicle speed, and v subj is the target vehicle speed after normalization;

[0015] The normalization method for the speed difference between the target vehicle and the environmental vehicle is:

[0016]

[0017] In the formula, Δv is the speed difference between the target vehicle and the environmental vehicle after normalization, and v sur is the environmental vehicle speed;

[0018] Since the range of the road surface friction coefficient is generally between 0 and 1, no normalization treatment is required.

[0019] Preferably, the first-layer input of the hierarchical fuzzy inference system in the step S2 includes the road surface friction coefficient μ and the normalized target vehicle speed v subj , and the output is the danger factor E; the second-layer input is the danger factor E and the environmental vehicle speed v sur , and the output is the steering decision D dec ; The construction steps of the hierarchical fuzzy inference system include:

[0020] Step a: Normalize the road surface friction coefficient μ, target vehicle speed v' subj and the speed difference Δv between the target vehicle and the environmental vehicle according to the method in the step S1, and the normalization range is set to [0, 1];

[0021] Step b: Establish the first membership function, second membership function, third membership function, fourth membership function, and fifth membership function. The first membership function and the second membership function are respectively used to transform the road surface friction coefficient μ and the normalized target vehicle speed v subj into the corresponding fuzzy variables and

[0022] Step c: Using the fuzzy variables and as the inputs of the first-layer fuzzy inference system, refine the general experience of vehicle driving to obtain the inference rule table of the risk factor E. According to the third membership function, infer the risk factor E as the fuzzy variable The general experience here is:

[0023] When is small and is large, take the large value;

[0024] When is small and is small, take the median value;

[0025] When is large and is small, take the small value;

[0026] When is large and is large, take the median value;

[0027] Step d: According to the fourth membership function, transform the speed difference Δv between the target vehicle and the environmental vehicle after normalization into the corresponding fuzzy variable Using and as the inputs of the second-layer fuzzy inference system, refine the general experience of vehicle driving to obtain the inference rule table of the steering decision D dec According to the fifth membership function, infer the steering decision D dec as the fuzzy variable The general experience here is:

[0028] When is small and is large, take the median value;

[0029] When is small and is small, take the small value;

[0030] When is large and is small, take the median value;

[0031] When is large and is large, take the large value;

[0032] Step e: Inverse normalize the fuzzy variable to The inverse normalization formula is as follows:

[0033]

[0034]

[0035] In the formula, is the decision distance, D s is the safety distance, a subj is the acceleration of the target vehicle, a sur is the acceleration of the surrounding vehicle, t x is the time delay of the braking system, d o is the shortest safety distance between vehicles;

[0036] Step e: In each iteration cycle, compare the distance d between the target vehicle and the surrounding vehicle with D dec When d ≤ D dec , the target vehicle performs a steering operation; otherwise, the target vehicle goes straight.

[0037] Preferably, in step S3, the steps of constructing the vehicle driving path based on the B-spline curve include:

[0038] Step A: Establish a vehicle path model based on the B-spline curve, and the expression is as follows:

[0039]

[0040] where B i,k (u) is the i-th k-th order B-spline basis function, related to the control point P i , k ≥ 1, τ is the independent variable;

[0041] The basis function has the following Cox-deBoor recurrence formula:

[0042]

[0043] In the formula, τ i is a set of continuously changing values of a non-decreasing sequence called the knot vector, with the first and last values defined as 0 and 1, and the sequence composed of τ i is as follows:

[0044] [τ 0 , τ 1 , …, τ k , τ k+1 , …, τ n , τ n+1 , …, τ n+k ;

[0045] Step B: According to the above recurrence formula, a cubic (4th-order) quasi-uniform B-spline curve can be generated by 6 control points (A, B, C, D, E, F).

[0046] Preferably, assume that the curvatures at the starting point and the ending point are 0, and ABC ∥ DEF ∥ the center line of the lane, and let AB = BC = DE = EF; the degrees of freedom of the 6 control points ABCDEF are reduced to 2, which are the distance k1 of point C relative to point A and the distance k2 of point O relative to point A.

[0047] Preferably, in the step S4, the steps of the particle swarm optimization method are as follows:

[0048] Step 1: Design a particle swarm optimization algorithm. The entire population consists of S particles moving in a D-dimensional search space; the position of the i-th particle is represented by a vector:

[0049]

[0050] where X min,m and X max,m are the upper and lower bounds of the m-th dimension respectively.

[0051] Its velocity update formula is:

[0052] V i (k + 1) = wV i (k) + c 1 r 1 [p best -X i (k + 1)] + c 2 r 2 [g best -X i (k)]

[0053] Its position update formula is:

[0054] X i (k + 1) = X i (k) + V i (k + 1)

[0055] where p best is the historical optimal position of this particle; g b4st is the optimal position of the entire particle swarm; k is the current iteration number; w is the inertial reference, which can be used to adjust the convergence direction; c 1 and c 2 are the local learning factor and the global learning factor respectively. The role of the c 1 learning factor is to adjust the step size of the particle moving towards p best , and c 2 is mainly responsible for adjusting the particle moving towards gbest The step sizes of learning, both of which are non - negative numbers; r 1 and r 2 are random numbers between [0, 1];

[0056] Step 2: Design the objective function. To measure the quality of these particles, the optimized objective function is defined by the path length and the average curvature of the path:

[0057] f = ω 1 f length + ω 2 f curvature

[0058]

[0059]

[0060] where f is the cost function, f length represents the path length, f curvature represents the average curvature of the path, ω 1 and ω 2 are weight coefficients; n is the number of all sampling points of the path to be optimized, K i is the curvature of each sampling point;

[0061] Step 3: Optimize and solve the undetermined parameters k1 and k2 in the vehicle path optimization model to obtain the steering path of the optimal path.

[0062] Preferably, the reference speed planning in step S5 includes the following steps:

[0063] Step E1: Set the corresponding fuzzy variables as S1, S2, S3, S4, S5, S6, S7, and S8 when the road surface friction coefficient μ is 0.25, 0.35, 0.45, 0.55, 0.65, 0.75, 0.85, and 0.95;

[0064] Step E2: Conduct sampling through random two - lane lane - changing experiments to obtain the optimal steering speed under different friction coefficients;

[0065] Step E3: Based on the above data, perform fuzzy processing through a Gaussian - shaped membership function to establish a mapping relationship between different road surface friction coefficients and the optimal steering speed.

[0066] The beneficial effects of the present invention:

[0067] In order to complete dynamic path planning under different road surface conditions and vehicle speeds and improve the lateral dynamic stability of the vehicle, the present invention proposes a dynamic path planning method for an autonomous vehicle. This method takes into account the road surface friction coefficient, the target vehicle, and the motion states of environmental vehicles, and uses methods such as hierarchical fuzzy inference and particle swarm optimization to calculate the driving behavior, driving path, and corresponding ideal speed that the vehicle should select at each moment. This planning method provides a new solution to the path planning problem in a complex driving environment and gives a new impetus to the development of this industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 FIG. is a flowchart of the implementation of a dynamic path planning method for an autonomous vehicle according to the present invention;

[0069] Figure 2 FIG. is a design block diagram of a dynamic path planning method for an autonomous vehicle according to the present invention;

[0070] Figure 3 is a simplified model for path optimization; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0071] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0072] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0073] See Figure 1 , this application example provides a dynamic path planning method for an autonomous vehicle, including the following steps:

[0074] Step S1, normalize driving environment factors such as road surface friction, target vehicle speed, and the speed difference between the target vehicle and environmental vehicles, and establish a driving environment model;

[0075] Step S2, combine the driving environment model, construct a hierarchical fuzzy inference system, dynamically update environmental parameters, and calculate the path type to be planned and the starting position of path planning at the current position. The path type includes executing a turn or going straight;

[0076] Step S3, construct a vehicle driving path based on the B-spline curve according to the vehicle dynamics characteristics, and obtain the undetermined parameters to be optimized;

[0077] Step S4: Considering two factors of distance and curvature, design an optimization method based on particle swarm to optimize and solve the undetermined parameters in the vehicle path optimization model, and obtain the optimal path;

[0078] Step S5: Use the method of fuzzy fitting to calculate the vehicle driving paths and corresponding reference speeds under different road surface conditions.

[0079] In the said step S1, the driving environment model is obtained by normalizing the road surface friction, the target vehicle speed, and the difference between the target vehicle speed and the environmental vehicle speed. Among them, the normalization method for the target vehicle speed is:

[0080]

[0081] In the formula, v min and v max are the minimum speed and the maximum speed of the target vehicle, which are set to 0 km / h and 120 km / h, v' subj is the target vehicle speed, and v subj is the target vehicle speed after normalization processing;

[0082] The normalization method for the difference between the target vehicle speed and the environmental vehicle speed is:

[0083]

[0084] In the formula, Δv is the difference between the target vehicle and the environmental vehicle speed after normalization processing, and v sur is the environmental vehicle speed;

[0085] Since the range of the road surface friction coefficient is generally between 0 and 1, no normalization processing is required.

[0086] The first-layer input of the hierarchical fuzzy inference system in the said step S2 includes the road surface friction coefficient μ and the normalized target vehicle speed v subj , and the output is the risk factor E; the second-layer input is the risk factor E and the environmental vehicle speed v sur , and the output is the steering decision D dec ; The construction steps of the hierarchical fuzzy inference system include:

[0087] Step a: Normalize the road surface friction coefficient μ, the target vehicle speed v' subj and the difference Δv between the target vehicle speed and the environmental vehicle speed according to the method in step S1, and the normalization range is set to [0, 1];

[0088] Step b: Establish the first membership function, the second membership function, the third membership function, the fourth membership function, and the fifth membership function. The first membership function and the second membership function are respectively used to transform the road surface friction coefficient μ and the normalized target vehicle speed v subj into the corresponding fuzzy variables and

[0089] Step c: Take the fuzzy variables and as the inputs of the first-layer fuzzy inference system, refine the general experience of vehicle driving, obtain the inference rule table of the risk factor E, and infer the risk factor E as the fuzzy variable The general experience here is:

[0090] When is small and is large, take the large value;

[0091] When is small and is small, take the median value;

[0092] When is large and is small, take the small value;

[0093] When is large and is large, take the median value;

[0094] Step d: According to the fourth membership function, transform the speed difference Δv between the normalized target vehicle and the environmental vehicle into the corresponding fuzzy variable Take and as the inputs of the second-layer fuzzy inference system, refine the general experience of vehicle driving, obtain the inference rule table of the steering decision D dec and infer the steering decision D dec as the fuzzy variable The general experience here is:

[0095] When is small and is large, take the median value;

[0096] When is small and is small, take the small value;

[0097] When Large and Hours, Take the median value;

[0098] When Large and Large, Take the large value;

[0099] Step e: Inverse normalize the fuzzy variable Inverse normalize to The inverse normalization formula is as follows:

[0100]

[0101]

[0102] Wherein, Is the decision-making distance, D s Is the safety distance, a subj Is the acceleration of the target vehicle, a sur Is the acceleration of the environmental vehicle, t x Is the time delay of the braking system, d o Is the shortest safety distance between vehicles;

[0103] Step e: In each iteration cycle, compare the distance d between the target vehicle and the environmental vehicle with D dec When d ≤ D dec , the target vehicle performs a steering operation, otherwise the target vehicle goes straight.

[0104] See Figure 3 , the B-spline curve is a parametric curve and a special case of the Bezier curve. The shape of the curve is determined by a set of control points connected in sequence. These control points form a control polygon, and the B-spline curve approximates the control polygon. The B-spline curve not only meets the requirement that the path curvature changes continuously, but also ensures that the curvature at the steering point and the merging point is zero, which is more suitable for driving requirements.

[0105] The steps for constructing a vehicle path optimization model based on the B-spline curve are as follows:

[0106] Step A: Establish a vehicle path model based on the B-spline curve, and the expression is as follows:

[0107]

[0108] Wherein, B i,k (u) is the i-th k-order B-spline basis function, related to the control point P i , k ≥ 1, τ is the independent variable.

[0109] The basis function has the following Cox-deBoor recurrence formula

[0110]

[0111] In the formula, τ i is a set of continuously varying values of a non-decreasing sequence called the knot vector, and the first and last values are generally defined as 0 and 1. τ i The sequence formed is as follows:

[0112] [τ 0 , τ 1 , …, τ k , τ k+1 , …, τ n , τ n+1 , …, τ n+k ;

[0113] Step B: According to the above recurrence formula, a cubic (4th order) quasi-uniform B-spline curve can be generated by 6 control points (A, B, C, D, E, F). To simplify the model and ensure that the curvatures at the starting and ending points are 0, the following considerations are made:

[0114] 1) The curvatures at the starting and ending points are zero, provided that three consecutive control points at the starting and ending points are on a straight line.

[0115] 2) To make the heading angles of the vehicle at the starting and ending points zero, according to the characteristics of the B-spline curve, the lines ABC and DEF need to be parallel to the center line.

[0116] 3) To ensure that the first half (left side of point O) and the second half (right side of point O) of the path are centrosymmetric, it is necessary to satisfy ABC = DEF.

[0117] In summary, let AB = BC = DE = EF.

[0118] Therefore, the degrees of freedom of the 6 control points ABCDEF are reduced to 2, which are the distance k of point C relative to point A 1 and the distance k of point O relative to point A 2 .

[0119] In step S4, the steps of the particle swarm optimization method are as follows:

[0120] Step 1: Design a particle swarm optimization algorithm. The entire population consists of S particles moving in a D-dimensional search space; the position of the i-th particle is represented by a vector as follows:

[0121]

[0122] where X min,m and X max,m are the upper and lower bounds of the m-th dimension respectively.

[0123] Its speed update formula is as follows:

[0124] V i (k + 1) = wV i (k) + c 1 r 1 [p best -X i (k + 1)] + c 2 r 2 [g best -X i (k)]

[0125] Its position update formula is as follows:

[0126] X i (k + 1) = X i (k) + V i (k + 1)

[0127] Among them, p best is the historical optimal position of the particle; g best is the optimal position of the entire particle swarm; k is the current iteration number; w is the inertial reference, which can be used to adjust the convergence direction; c 1 and c 2 are the local learning factor and the global learning factor respectively. The role of the c 1 learning factor is to adjust the step size of the particle to learn from p best The c 2 learning factor is mainly responsible for adjusting the step size of the particle to learn from g best Both are non - negative numbers; r 1 and r 2 are random numbers between [0, 1];

[0128] Step 2: Design the objective function. To measure the quality of these particles, the optimized objective function is defined by the path length and the path average curvature:

[0129] f = ω 1 f length + ω 2 f curvature

[0130]

[0131]

[0132] Among them, f is the cost function, f length represents the path length, f curvature represents the average curvature of the path, ω 1 and ω 2 are weight coefficients; n is the number of all sampling points of the path to be optimized, Ki is the curvature of each sampling point;

[0133] Step 3: Optimize and solve the undetermined parameters k1 and k2 in the vehicle path optimization model to obtain the steering path of the optimal path.

[0134] The reference speed planning in step S5 includes the following steps:

[0135] Step E1: Set the fuzzy variables corresponding to when the road surface friction coefficient μ is 0.25, 0.35, 0.45, 0.55, 0.65, 0.75, 0.85, and 0.95 as S1, S2, S3, S4, S5, S6, S7, and S8;

[0136] Step E2: Conduct sampling through random two-lane lane-changing experiments to obtain the optimal steering speed under different friction coefficients;

[0137] Step E3: Based on the above data, perform fuzzy processing through a Gaussian membership function to establish a mapping relationship between different road surface friction coefficients and the optimal steering speed.

[0138] The present invention designs a multi-dimensional dynamic state fusion inference method based on a hierarchical fuzzy inference system, infers the path type that needs to be planned at the current position according to information such as the main vehicle speed, the environmental vehicle speed, the road surface friction coefficient, and the dynamic distance between vehicles; when the vehicle should turn, according to the vehicle dynamics characteristics, construct a vehicle path optimization model based on the B-spline curve to obtain the undetermined parameters that need to be optimized; considering two factors of distance and curvature, design an optimization method based on particle swarm to optimize and solve the undetermined parameters in the vehicle path optimization model to obtain the optimal steering path; obtain the ideal steering speed corresponding to different road surface friction coefficients through experiments, and use the method of fuzzy fitting to calculate the reference speed corresponding to the above optimal path. Track the obtained path in CarSim, input the real-time data into the steering decision-making module for real-time inference, and complete the loop.

[0139] As described above, the path planning work carried out by the present invention can complete the driving decision-making of an autonomous vehicle in a dynamic environment, adapt to different road conditions and vehicle speeds, and improve the lateral dynamic stability of the vehicle during driving. Specifically, in order to complete dynamic path planning under different road conditions and vehicle speeds and improve the lateral dynamic stability of the vehicle, the present invention proposes a dynamic path planning method for an autonomous vehicle. This method takes into account the road surface friction coefficient, the target vehicle, and the motion states of environmental vehicles, and uses methods such as hierarchical fuzzy inference and particle swarm optimization to calculate the driving behavior, driving path, and corresponding ideal speed that the vehicle should choose at each moment. This planning method provides a new solution to the path planning problem in a complex driving environment and gives a new impetus to the development of this industry.

[0140] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for dynamic path planning of an autonomous driving vehicle, characterized in that: The method comprises the following steps: Step S1, normalizing driving environment factors to establish a driving environment model; Step S2: Combine the driving environment model to build a hierarchical fuzzy reasoning system, dynamically update the environmental parameters, and calculate the path type that needs to be planned at the current position, including turning or going straight; Step S3: constructing a vehicle driving path based on a B-spline curve according to the vehicle dynamics characteristics, and obtaining undetermined parameters that need to be optimized; Step S4, considering the two factors of distance and curvature, designing an optimization method based on particle swarm, optimizing and solving the undetermined parameters in the vehicle path optimization model, and obtaining the optimal path; Step S5, using a fuzzy fitting method to calculate a reference speed corresponding to the optimal path; The first level input of the hierarchical fuzzy inference system in step S2 includes the road friction coefficient μ and the normalized target vehicle speed v subj , the output is the risk factor E; the second layer input is the risk factor E and the environmental vehicle speed v sur , the output is the steering decision D dec .

2. The method for dynamic path planning of an autonomous driving vehicle according to claim 1, characterized in that: The driving environment factors include at least: road friction, target vehicle speed, and a speed difference between the target vehicle speed and the environment vehicle speed.

3. The method for dynamic path planning of an autonomous driving vehicle according to claim 2, characterized in that: In step S1, the driving environment model is obtained by normalizing the road friction, the target vehicle speed, and the speed difference between the target vehicle speed and the environment vehicle speed, wherein the target vehicle speed normalization method is: In the formula, v min and v max are the minimum and maximum speeds of the target vehicle, which are set to 0km / h and 120km / h, v' subj is the target vehicle speed, v subj is the normalized target vehicle speed; The method for normalizing the speed difference between the target vehicle speed and the environment vehicle speed is: Where Δv is the normalized speed difference between the target vehicle and the surrounding vehicle, v sur is the ambient vehicle speed; The road friction coefficient μ ranges between 0-1.

4. The method for dynamic path planning of an autonomous driving vehicle according to claim 3, characterized in that: The steps of constructing the hierarchical fuzzy reasoning system include: Step a: According to the method in step S1, the road friction coefficient μ and the target vehicle speed v' subj The speed difference Δv between the target vehicle and the environment vehicle is normalized, and the normalized range is set to [0,1]; Step b: Establish a first membership function, a second membership function, a third membership function, a fourth membership function and a fifth membership function, wherein the first membership function and the second membership function are used to convert the road friction coefficient μ, the normalized target vehicle speed v subj Transformed into corresponding fuzzy variables and Step c: Fuzzy variables and As the input of the first-level fuzzy inference system, the general experience of vehicle driving is refined to obtain the inference rule table of the risk factor E. According to the third membership function, the risk factor E is inferred as a fuzzy variable Step d: According to the fourth membership function, the normalized speed difference Δv between the target vehicle and the environment vehicle is converted into the corresponding fuzzy variable by and The fuzzy variables are used as the input of the second-level fuzzy inference system, and the general experience of vehicle driving is refined to obtain the steering decision D dec The inference rule table will turn to decision D according to the fifth membership function dec Reasoning for fuzzy variables Step e: Fuzzy variables Denormalize to The inverse normalization formula is as follows: In the formula, is the decision distance, D s is the safe distance, a subj is the acceleration of the target vehicle, a sur is the acceleration of the ambient vehicle, t x is the time delay of the braking system, d o is the shortest safe distance between vehicles; Step e: In each iteration cycle, the distance d between the target vehicle and the environment vehicle is dec For comparison, when d≤D dec , the target vehicle performs a steering operation, otherwise the target vehicle goes straight.

5. The method for dynamic path planning of an autonomous driving vehicle according to claim 4, characterized in that: In step S3, the step of constructing the vehicle driving path based on the B-spline curve includes: Step A: Establish a vehicle path model based on B-spline curve, the expression is as follows: Among them, B i,k (u) is the i-th k-order B-spline basis function, and the control point P i ,k≥1,τ is the independent variable; The basis function has the following Cox-deBoor recursion: In the formula, τ i It is a set of continuously changing values ​​in a non-decreasing sequence called node vectors. The first and last values ​​are generally defined as 0 and 1. i The sequence is as follows: [τ0,τ1,…,τ k ,t k+1 ,…,t n ,t n+1 ,…,t n+k ]; Step B: According to the above recursive formula, a cubic quasi-uniform B-spline curve can be generated by 6 control points (A, B, C, D, E, F).

6. The method for dynamic path planning of an autonomous driving vehicle according to claim 5, characterized in that: Assume that the curvature of the starting point and the end point is 0, and the line ABC|DEF|lane center line, let AB=BC=DE=EF; the degree of freedom of the 6 control points ABCDEF is reduced to 2, which are the distance k1 of point C relative to point A and the distance k2 of point O relative to point A.

7. The method for dynamic path planning of an autonomous driving vehicle according to claim 6, characterized in that: In step S4, the particle swarm optimization method steps are as follows: Step 1: Design a particle swarm optimization algorithm. The entire swarm consists of S particles moving in a D-dimensional search space. The position of the i-th particle is represented by a vector: Among them, X min,m and X max,m are the upper and lower bounds of the mth dimension respectively; The speed update formula is: V i (k+1)=wV i (k)+c1r1[p best -X i (k+1)]+c2r2[g best -X i (k)] Its position update formula is: X i (k+1)=X i (k)+V i (k+1) Among them, p best is the historical optimal position of the particle; g best is the optimal position of the entire particle swarm; k is the current number of iterations; w is the inertial reference, which can be used to adjust the convergence direction; c1 and c2 are local learning factors and global learning factors, respectively. The role of c1 learning factor is to adjust the particle to p best The learning step size, c2 is mainly responsible for adjusting the particle to g best The learning step size, both are non-negative numbers; r1 and r2 are random numbers between [0,1]; Step 2: Design the objective function. In order to measure the quality of these particles, the optimized objective function is defined by the path length and the average curvature of the path: f=ω1f length +ω2f curvature Among them, f is the cost function, f length represents the path length, f curvature represents the average curvature of the path, ω1 and ω2 are weight coefficients; n is the number of sampling points that need to be optimized, K i is the curvature of each sampling point; Step 3: Optimize and solve the unknown parameters k1 and k2 in the vehicle path optimization model to obtain the steering path of the optimal path.

8. The method for dynamic path planning of an autonomous driving vehicle according to claim 7, characterized in that: The reference speed planning in step S5 comprises the following steps: Step E1: setting the corresponding fuzzy variables to S1, S2, S3, S4, S5, S6, S7 and S8 when the road friction coefficient μ is 0.25, 0.35, 0.45, 0.55, 0.65, 0.75, 0.85 and 0.95; Step E2: sampling through a random two-lane lane change experiment to obtain the optimal steering speed under different friction coefficients; Step E3: Based on the above data, fuzzy processing is performed through a Gaussian membership function to establish a mapping relationship between different road friction coefficients and optimal steering speeds.

Images