Trajectory generation and risk assessment method for urban automatic driving

Abstract

The invention relates to a track generation and risk assessment method for urban automatic driving, and belongs to the field of intelligent driving and vehicle track planning, and the method comprises the following steps: obtaining a global reference path based on a road topology structure; generating an initial track cluster in the curve coordinate system; performing feasibility judgment and pruning on the candidate tracks; a multi-target dynamic risk assessment mechanism is constructed, and the self-vehicle dynamic risk brought by the trajectory is quantified; constructing a deterministic risk quantification model fusing the transverse deviation, and quantifying the deterministic risk of the trajectory; reasoning a future space-time occupation area of each dynamic obstacle, constructing an uncertainty risk quantification model, and determining a trajectory uncertainty risk; performing one-by-one risk assessment on the effective trajectories, and calculating the comprehensive risk of each trajectory in combination with a self-vehicle dynamic risk, certainty and uncertainty risk quantification model; and selecting an optimal track output as an execution track based on the comprehensive risk.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent driving and vehicle trajectory planning, and relates to a trajectory generation and risk assessment method for urban autonomous driving. Background Art

[0002] With the gradual commercialization of autonomous driving technology in highways and simple scenarios, its technological focus is gradually expanding to highly complex environments such as urban roads. Urban roads, characterized by a high density of traffic participants, diverse scene topologies, and frequent dynamic interactions, pose greater challenges to autonomous driving systems in terms of perception, decision-making, and planning. Ensuring safe and efficient passage of autonomous vehicles in urban environments is one of the key bottlenecks in the current commercialization of intelligent driving systems. In this context, trajectory planning and risk assessment, as core components of autonomous driving decision-making systems, directly determine the rationality, safety, and passenger acceptance of vehicle driving behavior, and have become key areas for technological breakthroughs in urban autonomous driving systems.

[0003] The essence of trajectory planning lies in generating a feasible trajectory to a target area based on the vehicle's current position, sensory information, and road topology, while satisfying dynamic constraints, safety constraints, and behavioral preferences. To achieve efficient trajectory generation and selection, current mainstream approaches include traditional trajectory cluster generation methods based on spatial or control quantity sampling, and intelligent approaches that utilize deep learning models to output trajectories end-to-end. Traditional methods often utilize the Frenet coordinate system to construct a road-aligned coordinate system and obtain candidate trajectory clusters by fitting polynomial curves to spatial sampling. This approach strikes a good balance between computational efficiency and trajectory expressiveness, but often fails to consider vehicle kinematic constraints. On the other hand, with the development of data-driven technologies, end-to-end trajectory prediction and planning models have gradually emerged. These models directly generate future multimodal trajectories from sensory and navigation information, improving their ability to handle complex scenarios. However, these methods often lack explicit control and interpretability over the trajectory generation process, making it difficult to effectively evaluate and verify the generated results, posing potential risks.

[0004] Risk factors in urban road scenarios come from a wide range of sources. These include not only constraints from the surrounding static environment, such as road boundaries, lane markings, and static obstacles, but also the uncertain behavior of dynamic obstacles, the vehicle's own dynamic constraints, and the dynamic risks induced by the trajectory itself. Therefore, a risk modeling and quantification mechanism that integrates multiple information sources is urgently needed. This mechanism can not only model the dynamic risks posed by the trajectory, but also describe the deterministic risks posed by static geometric conflicts. Furthermore, it can combine trajectory prediction with probabilistic reasoning to model the potential collision uncertainty in dynamic interactions. Only by incorporating multi-source risk factors into a trajectory assessment system and quantitatively assessing trajectory risks during the planning phase can the safety and robustness of urban autonomous driving systems in complex dynamic environments be effectively guaranteed. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a trajectory generation and risk assessment method for urban autonomous driving. By adopting a horizontal and vertical decoupled sampling strategy in a curvilinear coordinate system to generate candidate trajectory clusters, and combining them with kinematic consistency constraints for screening, the feasibility and road adaptability of the trajectories are improved. The uncertainty factors of the vehicle's dynamic, static obstacles, and dynamic obstacles are integrated to construct a multidimensional risk quantification model to comprehensively evaluate the safety of each trajectory. Finally, the optimal trajectory output is selected based on a multi-source risk weighting strategy, effectively improving the trajectory planning performance of autonomous vehicles in complex urban environments.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A trajectory generation and risk assessment method for urban autonomous driving includes the following steps:

[0008] S1: Based on the vehicle's current position, the local road curve coordinate system, and perception information, the starting point and target point are set, and the global reference path is obtained based on the road topology.

[0009] S2: Generate an initial trajectory cluster using a horizontal and vertical decoupled sampling strategy in the curvilinear coordinate system to obtain multiple candidate trajectories;

[0010] S3: Combined with the kinematic consistency constraints, the feasibility of the candidate trajectories is judged and pruned, and the valid trajectories that meet the motion constraints are retained to form a trajectory set;

[0011] S4: Considering safety and efficiency comprehensively, a multi-objective dynamic risk assessment mechanism is constructed to quantify the dynamic risk of the ego vehicle brought by the trajectory;

[0012] S5: Based on the static obstacle information, road boundaries, and lane centerlines within the perception range, a deterministic risk quantification model is constructed that incorporates lateral deviation to quantify the trajectory deterministic risk.

[0013] S6: Based on the trajectory prediction results of dynamic obstacles and their covariance propagation, the future spatiotemporal occupied areas of each dynamic obstacle are inferred, and an uncertainty risk quantification model is constructed to determine the trajectory uncertainty risk;

[0014] S7: Perform risk assessment on each valid trajectory obtained in step S3, and calculate the comprehensive risk of each trajectory by combining the dynamic risk of the vehicle and the deterministic and uncertain risk quantification models;

[0015] S8: Select the optimal trajectory output as the execution trajectory based on the comprehensive risk.

[0016] Furthermore, in step S1, obtaining a global reference path based on the road topology structure includes: calculating an optimal path consisting of a series of key nodes and curve segments based on lane connectivity, combined with the current position and the target point, for constructing a curve coordinate system.

[0017] Furthermore, in step S2, the horizontal and vertical decoupling sampling strategy includes:

[0018] S21: In the planning time domain T, sample the planning time domain to obtain target points in different time domains, and use the fifth-order polynomial in the longitudinal direction to obtain the fitting curve of the longitudinal position:

[0019] s(t)=a0+a1t 1 +a2t 2 +a3t 3 +a4t 4 +a5t 5

[0020] Where a0~a5 are the coefficients of the fifth-order polynomial, i=0,1,…,5, and t is a moment in the planning time domain;

[0021] S22: For each longitudinal polynomial trajectory, the lateral direction is centered at the target offset d=0, and different lateral final states are obtained by sampling a small range, and a fitting curve of the lateral position is generated based on the cubic polynomial;

[0022] d(t)=b0+b1t+b2t 2 +b3t 3

[0023] Where b0~b3 are cubic polynomial coefficients;

[0024] S23: Construct a complete sampling trajectory cluster based on the fitting curves of different time, horizontal and vertical positions

[0025] Furthermore, in step S3, the kinematic consistency constraint includes:

[0026] Vehicle longitudinal velocity continuity constraints, curvature boundary constraints, yaw rate and its derivative boundary constraints are used to screen out candidate trajectories with non-smooth dynamic behavior or that do not meet the actual capabilities of the vehicle, and obtain pruned trajectory clusters for risk assessment

[0027] Furthermore, the multi-objective dynamic risk assessment mechanism is constructed in step S4 to evaluate the dynamic risk of the vehicle brought by the tracking trajectory τ The specific calculation includes the following steps:

[0028] S41: Obtain the acceleration value at a specific discrete time step and calculate the vehicle acceleration risk R caused by the trajectory through weighted calculation. a :

[0029]

[0030] Where N is the maximum time discrete sampling step, λ a represents the acceleration risk weight coefficient, a i Indicates the acceleration value of the trajectory point at time i;

[0031] S42: The reference speed v of the vehicle on the map is known ref , quantized trajectory velocity and reference velocity v ref The degree of deviation is used to calculate the vehicle speed risk:

[0032]

[0033] Where λ v1 ~λ v3 Represents the speed risk weight coefficient of different trajectory points, v i represents the speed at time i, v N represents the speed at time N, v N / 2 represents the speed at time N / 2;

[0034] S43: Reference heading θ based on the nearest point on the reference path ref,i , quantifying the risk R caused by the heading deviation between the trajectory and the curved reference path y :

[0035]

[0036] Where λ y1 ,λ y2 represents the heading risk weight coefficient, θ i represents the heading of the trajectory point at time i, θ N Represents the heading of the trajectory point at time N, θ ref,i Indicates the heading of the reference path point corresponding to the trajectory point at time i, θ ref,NIndicates the heading of the reference path point corresponding to the trajectory point at time N;

[0037] S44: Based on the determined risks, the dynamic risk of the vehicle in trajectory τ is obtained by weighting

[0038]

[0039] Where, ω a ,ω v ,ω y They represent the weights of acceleration risk, speed risk and heading risk respectively.

[0040] Furthermore, the deterministic risk quantification model integrating the lateral deviation is constructed in step S5 to evaluate the risk of trajectory τ caused by environmental deterministic factors. The specific calculation includes the following steps:

[0041] S51: Set the vehicle’s sensing range to the current position of the vehicle (x e ,y e ,θ e ) as the center, with a circular area of perception radius R, and a resolution of Δr, the perception range is rasterized to obtain a grid point set in the Cartesian coordinate system.

[0042] in, represents a set of non-negative integers, i and j represent the row and column indices of the grid points;

[0043] Filter the following grid point sets that meet the perception range

[0044]

[0045] S52: For each grid point within the perception range Quantify the risk r due to three factors: static obstacles, road topology, and lateral deviation from the reference path static,i,j , r topo,i,j , r path,i,j :

[0046]

[0047] Where, Represents the polygonal area set occupied by static obstacles, P k Represents a polygonal area of the map;

[0048]

[0049] Where, Represents a set of lane polygon areas in road topology;

[0050]

[0051] Where, d max Indicates the maximum lateral deviation distance threshold, d i,j Indicates the shortest distance between the grid point and the reference path;

[0052] S53: Taking into account the static obstacles in the environment, the road topology, and the deviation between the spatial point and the reference path, a normalized deterministic risk is established for each grid point within the sensing range:

[0053] R i,j =min((r static,i,j +r topo,i,j +r path,i,j ),1)

[0054] S54: Obtain the deterministic risk of the trajectory by querying the risk value of each trajectory point within the perception range on the trajectory τ and then accumulating it

[0055] Furthermore, the trajectory uncertainty risk quantification model is constructed in step S6 to evaluate the uncertainty risk of the trajectory τ caused by dynamic obstacles. The specific calculation includes the following steps:

[0056] S61: Construct the future space occupation area based on the known obstacle geometry and the predicted information at the future time t Expressed as:

[0057]

[0058] Where, L i ,W i Indicates the length and width of the obstacle, Indicates the heading and geometric center coordinates of the obstacle;

[0059] S62: Construct an elliptical region ε of equal probability based on the covariance matrix of the prediction results t , represents the possible location area of obstacles within the confidence range, and the center of the elliptical area is The major axis w and minor axis h of the elliptical area are determined by the eigenvalues λ1 and λ2 of the covariance matrix:

[0060]

[0061] Where α is the scaling factor;

[0062] S63: Quantify the uncertainty occupied area of the obstacle vehicle at time t The risk brought by using the covariance matrix in the forecast information Quantify the uncertainty risk of the occupied area. For the trajectory point in the trajectory τ, the uncertainty risk value at time t is

[0063]

[0064] Where, represents the uncertainty risk quantitative index of the occupied area at time t, σ xy Represents the covariance of the prediction results in the x and y directions, σ x Indicates the variance of the prediction result in the x direction, σ y Indicates the variance of the prediction result in the y direction;

[0065] S64: By correlating the spatiotemporal elements of each trajectory point on the trajectory τ and the predicted occupied area, the uncertainty risk of each point on the trajectory is accumulated to obtain the uncertainty risk of the trajectory

[0066]

[0067] Furthermore, in step S7, the comprehensive risk of trajectory τ for:

[0068]

[0069] Where η dyn ,η det ,η unc They represent the weight coefficients of the vehicle's dynamic risk, deterministic risk, and uncertainty risk respectively.

[0070] Further, step S8 is based on the comprehensive risk of each trajectory obtained in S7 Select Minimum trajectory τ * Output as the optimal trajectory.

[0071] The beneficial effects of the present invention are as follows: The present invention aims to propose a trajectory generation and risk assessment method for urban autonomous driving. First, by combining horizontal and vertical decoupled sampling in a curvilinear coordinate system with kinematic consistency constraints, a candidate trajectory cluster with strong coverage and satisfying the vehicle's motion characteristics is generated, thereby improving the feasibility and diversity of the trajectory, reducing the number of alternative trajectories and improving the evaluation efficiency; then, a unified multi-index trajectory evaluation mechanism is constructed, integrating the dynamic risk of the vehicle, the deterministic risk brought by static obstacles, and the uncertain risk caused by dynamic obstacles, to achieve multi-dimensional and fine-grained risk modeling and enhance the accuracy of risk assessment; a comprehensive evaluation of alternative trajectories is performed through a multi-source risk weighting mechanism, dynamically adapting to changes in the complex urban environment, and improving the trajectory safety, smoothness, and efficiency of autonomous vehicles in urban areas.

[0072] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0074] Figure 1 This is a framework diagram of the trajectory generation and risk assessment method for urban autonomous driving according to the present invention;

[0075] Figure 2 Map information and global reference path diagram;

[0076] Figure 3 A complete trajectory cluster for horizontal and vertical decoupled sampling;

[0077] Figure 4 This is a raster diagram of the vehicle's perception range;

[0078] Figure 5 The gridded deterministic risk distribution at a certain moment;

[0079] Figure 6 is the position and area of the equal probability ellipse of the prediction result;

[0080] Figure 7 The temporal and spatial occupied areas and risk value distribution of dynamic obstacle vehicles, where (a) is a 2D angle diagram and (b) is a 3D angle diagram;

[0081] Figure 8 Schematic diagram of the spatiotemporal correlation mechanism of trajectory uncertainty risk. DETAILED DESCRIPTION

[0082] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0083] It should be noted that the illustrations provided in the following embodiments are merely schematic illustrations of the basic concept of the present invention. Therefore, the illustrations only show components related to the present invention and are not drawn according to the number, shape, and size of components in actual implementation. In actual implementation, the type, quantity, and proportion of each component may be changed arbitrarily, and the component layout may also be more complex.

[0084] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the embodiments of the present invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring the embodiments of the present invention.

[0085] Example 1:

[0086] The present invention provides a trajectory generation and risk assessment method for urban autonomous driving. Figure 1 As shown, the following steps are included:

[0087] Step 1: Based on the vehicle's current position, local road curve coordinate system and perception information, set the target position, obtain map information and global reference path based on the road topology, such as Figure 2 As shown;

[0088] Step 2: Generate an initial trajectory cluster using a horizontal and vertical decoupled sampling strategy in the curvilinear coordinate system to obtain multiple candidate trajectories. This step is as follows:

[0089] 1) Within the planning time domain T, sample the planning time domain to obtain target points in different time domains, and use a quintic polynomial in the longitudinal direction to obtain the fitting curve of the longitudinal position:

[0090] s(t)=a0+a1t 1 +a2t 2 +a3t 3 +a4t 4+a5t 5

[0091] Where a0~a5 are the coefficients of the fifth-order polynomial, i=0,1,…,5, and t is a moment in the planning time domain;

[0092] 2) For each longitudinal polynomial trajectory, the lateral direction is centered at the target offset d = 0, and different lateral final states are obtained by sampling a small range. A fitting curve of the lateral position is generated based on the cubic polynomial;

[0093] d(t)=b0+b1t+b2t 2 +b3t 3

[0094] Where b0~b3 are cubic polynomial coefficients;

[0095] 3) Construct a complete sampling trajectory cluster based on the fitting curves of different time, horizontal and vertical positions like Figure 3 shown.

[0096] Step 3: Vehicle longitudinal velocity continuity constraints, curvature boundary constraints, yaw rate and its derivative boundary constraints are used to screen out candidate trajectories with non-smooth dynamic behavior or that do not meet the actual capabilities of the vehicle, and obtain pruned trajectory clusters for risk assessment.

[0097] Step 4: Construct a multi-objective dynamic risk assessment mechanism to evaluate the dynamic risk of the vehicle in the tracking trajectory τ The steps are as follows:

[0098] 1) Obtain the acceleration value at a specific discrete time step and calculate the vehicle acceleration risk R caused by the trajectory by weighted calculation a :

[0099]

[0100] Where N is the maximum time discrete sampling step, λ a represents the acceleration risk weight coefficient, a i Indicates the acceleration value of the trajectory point at time i;

[0101] 2) The reference speed v of the vehicle on the map is known ref , quantized trajectory velocity and reference velocity v ref The degree of deviation is used to calculate the vehicle speed risk:

[0102]

[0103] Where λ v1 ~λ v3Represents the speed risk weight coefficient of different trajectory points, v i represents the speed at time i, v N represents the speed at time N, v N / 2 represents the speed at time N / 2;

[0104] 3) Reference heading θ based on the nearest point on the reference path ref,i , quantifying the risk R caused by the heading deviation between the trajectory and the curved reference path y :

[0105]

[0106] Where λ y1 ,λ y2 represents the heading risk weight coefficient, θ i represents the heading of the trajectory point at time i, θ N Represents the heading of the trajectory point at time N, θ ref,i Indicates the heading of the reference path point corresponding to the trajectory point at time i, θ ref,N Indicates the heading of the reference path point corresponding to the trajectory point at time N;

[0107] 4) Determine each risk according to the above steps and obtain the dynamic risk of the vehicle in trajectory τ by weighting

[0108]

[0109] Where, ω a ,ω v ,ω y They represent the weights of acceleration risk, speed risk and heading risk respectively.

[0110] Step 5: Construct a trajectory deterministic risk quantification model to evaluate the risk of trajectory τ caused by environmental deterministic factors. The steps are as follows:

[0111] 1) Set the vehicle’s sensing range to the current position of the vehicle (x e ,y e ,θ e ) as the center and the circular area with a sensing radius of R. To reduce the amount of calculation while ensuring accuracy, the resolution is defined as Δr, and the sensing range is rasterized to obtain a set of grid points in the Cartesian coordinate system.

[0112]

[0113] in, Represents a set of non-negative integers, i, j represents the row and column indexes of the grid points. In order to ensure that only grid points within the perception range are considered, the following grid point sets that meet the perception range are screened. like Figure 4 As shown:

[0114]

[0115] 2) For each grid point within the perception range Quantify the risk r due to three factors: static obstacles, road topology, and lateral deviation from the reference path static,i,j , r topo,i,j , r path,i,j :

[0116]

[0117] Where, Represents the polygonal area set occupied by static obstacles, P k Represents a polygonal area of the map;

[0118]

[0119] Where, Represents a set of lane polygon areas in road topology;

[0120]

[0121] Where, d max Indicates the maximum lateral deviation distance threshold, d i,j Indicates the shortest distance between the grid point and the reference path;

[0122] 3) Taking into account static obstacles in the environment, road topology, and the deviation between the spatial point and the reference path, a normalized deterministic risk is established for each grid point within the sensing range:

[0123] R i,j =min((r static,i,j +r topo,i,j +r path,i,j ),1)

[0124] Get the gridded deterministic risk distribution at the current moment, such as Figure 5 As shown;

[0125] 4) By querying the risk value of each trajectory point on the trajectory τ within the perception range and then accumulating it, the deterministic risk of the trajectory is obtained

[0126] Step 6: Construct a trajectory uncertainty risk quantification model to evaluate the uncertainty risk of trajectory τ caused by dynamic obstacles The steps are as follows:

[0127] 1) Based on the known obstacle geometry and the predicted information at the future time t, construct the future space occupation area Expressed as:

[0128]

[0129] Where, L i ,W i Indicates the length and width of the obstacle, Indicates the heading and geometric center coordinates of the obstacle;

[0130] 2) Construct an elliptical region of equal probability ε based on the covariance matrix of the prediction results t , represents the possible location area of obstacles within the confidence range. The center of the elliptical area is The major axis w and minor axis h of the elliptical area are determined by the eigenvalues λ1 and λ2 of the covariance matrix, as follows: Figure 6 As shown:

[0131]

[0132] Where α is the scaling factor;

[0133] 3) Quantify the uncertainty occupied area of the obstacle vehicle at time t The risk brought by using the covariance matrix in the forecast information Quantify the uncertainty risk of the occupied area. For the trajectory point in the trajectory τ, the uncertainty risk value at time t is

[0134]

[0135] Where, represents the uncertainty risk quantitative index of the occupied area at time t, σ xy Represents the covariance of the prediction results in the x and y directions, σ x Indicates the variance of the prediction result in the x direction, σ y Indicates the variance of the prediction result in the y direction. Figure 7 As shown in (a) and (b), the future spatiotemporal occupancy area and risk value distribution of a dynamic obstacle vehicle are displayed from 2D and 3D perspectives;

[0136] 4) By correlating the spatiotemporal elements of each trajectory point on the trajectory τ and the predicted occupied area (such as Figure 8 As shown), the uncertainty risk of each point on the trajectory is accumulated to obtain the uncertainty risk of the trajectory

[0137]

[0138] Step 7: Construct a multi-source risk weighting mechanism to obtain the trajectory τ comprehensive risk

[0139]

[0140] Where η dyn ,η det ,η unc They represent the weight coefficients of the vehicle’s dynamic risk, deterministic risk, and uncertainty risk respectively;

[0141] Step 8: Based on the comprehensive risk of each trajectory obtained in step 7 Select Minimum trajectory τ * Output as the optimal trajectory.

[0142] Example 2:

[0143] An electronic device comprising a memory and a processor;

[0144] The memory is used to store computer programs;

[0145] The processor is configured to implement the method described in Example 1 when executing the computer program.

[0146] Example 3:

[0147] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in Example 1 is implemented.

[0148] Example 4:

[0149] A computer program product includes a computer program, which implements the method described in embodiment 1 when executed by a processor.

[0150] In the above embodiments, references to "this embodiment" in the specification indicate that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in at least some embodiments, but not necessarily all embodiments. Multiple occurrences of "this embodiment" do not necessarily refer to the same embodiment.

[0151] In the above embodiments, although the invention has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those skilled in the art based on the foregoing description. For example, other memory structures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed. The embodiments of the present invention are intended to encompass all such alternatives, modifications, and variations that fall within the broad scope of the appended claims.

[0152] Regarding the computer-readable storage medium in this embodiment, those skilled in the art will appreciate that all or part of the steps in the aforementioned method embodiments can be implemented using hardware associated with the computer program. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps in the aforementioned method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0153] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication with each other. The memory is used to store computer programs, the communication interface is used for communication, and the processor and the transceiver are used to run computer programs so that the electronic terminal executes the various steps of the above method.

[0154] In this embodiment, the memory may include a random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage.

[0155] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0156] The present invention can be used in a wide variety of general-purpose or special-purpose computing system environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments that include any of the above.

[0157] The present invention may be described in the general context of computer-executable instructions, such as program modules, executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media, including storage devices.

[0158] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A trajectory generation and risk assessment method for autonomous driving in urban areas, characterized by: The following steps are involved: S1: Based on the vehicle's current position, the local road curve coordinate system, and perception information, the starting point and target point are set, and the global reference path is obtained based on the road topology. S2: Generate an initial trajectory cluster using a horizontal and vertical decoupled sampling strategy in the curvilinear coordinate system to obtain multiple candidate trajectories; S3: Combined with the kinematic consistency constraints, the feasibility of the candidate trajectories is judged and pruned, and the valid trajectories that meet the motion constraints are retained to form a trajectory set; S4: Considering safety and efficiency comprehensively, a multi-objective dynamic risk assessment mechanism is constructed to quantify the dynamic risk of the ego vehicle brought by the trajectory; S5: Based on the static obstacle information, road boundaries, and lane centerlines within the perception range, a deterministic risk quantification model is constructed that incorporates lateral deviation to quantify the trajectory deterministic risk. S6: Based on the trajectory prediction results of dynamic obstacles and their covariance propagation, the future spatiotemporal occupied areas of each dynamic obstacle are inferred, and an uncertainty risk quantification model is constructed to determine the trajectory uncertainty risk; S7: Perform risk assessment on each valid trajectory obtained in step S3, and calculate the comprehensive risk of each trajectory by combining the dynamic risk of the vehicle and the deterministic and uncertain risk quantification models; S8: Select the optimal trajectory output as the execution trajectory based on the comprehensive risk.

2. The trajectory generation and risk assessment method for urban autonomous driving according to claim 1, characterized in that: In step S1, obtaining a global reference path based on the road topology structure includes: calculating an optimal path consisting of a series of key nodes and curve segments based on lane connectivity, combined with the current position and the target point, for constructing a curve coordinate system.

3. The trajectory generation and risk assessment method for urban autonomous driving according to claim 1, characterized in that: In step S2, the horizontal and vertical decoupling sampling strategy includes: S21: In the planning time domain T, sample the planning time domain to obtain target points in different time domains, and use the fifth-order polynomial in the longitudinal direction to obtain the fitting curve of the longitudinal position: s(t)=a0+a1t 1 +a2t 2 +a3t 3 +a4t 4 +a5t 5 Where a0~a5 are the coefficients of the fifth-order polynomial, i=0,1,…,5, and t is a certain moment in the planning time domain; S22: For each longitudinal polynomial trajectory, the lateral direction is centered at the target offset d=0, and different lateral final states are obtained by sampling a small range, and a fitting curve of the lateral position is generated based on the cubic polynomial; d(t)=b0+b1t+b2t 2 +b3t 3 Where b0~b3 are cubic polynomial coefficients; S23: Construct a complete sampling trajectory cluster based on the fitting curves of different time, horizontal and vertical positions 4. The trajectory generation and risk assessment method for urban autonomous driving according to claim 1, characterized in that: In step S3, the kinematic consistency constraints include: Vehicle longitudinal velocity continuity constraints, curvature boundary constraints, yaw rate and its derivative boundary constraints are used to screen out candidate trajectories with non-smooth dynamic behavior or that do not meet the actual capabilities of the vehicle, and obtain pruned trajectory clusters for risk assessment 5. The trajectory generation and risk assessment method for urban autonomous driving according to claim 1, characterized in that: Step S4 constructs a multi-objective dynamic risk assessment mechanism to evaluate the dynamic risk of the vehicle in the tracking trajectory τ. The specific calculation includes the following steps: S41: Obtain the acceleration value at a specific discrete time step and calculate the vehicle acceleration risk R caused by the trajectory through weighted calculation. a : Where N is the maximum time discrete sampling step, λ a represents the acceleration risk weight coefficient, a i Indicates the acceleration value of the trajectory point at time i; S42: The reference speed v of the vehicle on the map is known ref , quantized trajectory velocity and reference velocity v ref The degree of deviation is used to calculate the vehicle speed risk: Where λ v1 ~λ v3 Represents the speed risk weight coefficient of different trajectory points, v i represents the speed at time i, v N represents the speed at time N, v N / 2 represents the speed at time N / 2; S43: Reference heading θ based on the nearest point on the reference path ref,i , quantifying the risk R caused by the heading deviation between the trajectory and the curved reference path y : Where λ y1 ,λ y2 represents the heading risk weight coefficient, θ i represents the heading of the trajectory point at time i, θ N Represents the heading of the trajectory point at time N, θ ref,i Indicates the heading of the reference path point corresponding to the trajectory point at time i, θ ref,N Indicates the heading of the reference path point corresponding to the trajectory point at time N; S44: Based on the determined risks, the dynamic risk of the vehicle in trajectory τ is obtained by weighting Where, ω a ,ω v ,ω y They represent the weights of acceleration risk, speed risk and heading risk respectively.

6. The trajectory generation and risk assessment method for urban autonomous driving according to claim 5, characterized in that: Step S5 constructs a deterministic risk quantification model that integrates lateral deviations to evaluate the risk of trajectory τ caused by environmental deterministic factors. Its specific calculation The method comprises the following steps: S51: Set the vehicle’s sensing range to the current position of the vehicle (x e ,y e ,θ e ) as the center, with a circular area of perception radius R, and a resolution of Δr, the perception range is rasterized to obtain a grid point set in the Cartesian coordinate system. in, represents a set of non-negative integers, i and j represent the row and column indices of the grid points; Filter the following grid point sets that meet the perception range S52: For each grid point within the perception range Quantify the risk r due to three factors: static obstacles, road topology, and lateral deviation from the reference path static,i,j , r topo,i,j , r path,i,j : Where, Represents the polygonal area set occupied by static obstacles, P k Represents a polygonal area of the map; Where, Represents a collection of lane polygon areas in road topology; Where, d max Indicates the maximum lateral deviation distance threshold, d i,j Indicates the shortest distance between the grid point and the reference path; S53: Taking into account the static obstacles in the environment, the road topology, and the deviation between the spatial point and the reference path, a normalized deterministic risk is established for each grid point within the sensing range: R i,j =min((r static,i,j +r topo,i,j +r path,i,j ),1) S54: Obtain the deterministic risk of the trajectory by querying the risk value of each trajectory point within the perception range on the trajectory τ and then accumulating it 7. The trajectory generation and risk assessment method for urban autonomous driving according to claim 6, characterized in that: Step S6 is to construct a trajectory uncertainty risk quantification model to evaluate the uncertainty risk of trajectory τ caused by dynamic obstacles. The specific calculation includes the following steps: S61: Construct the future space occupation area based on the known obstacle geometry and the predicted information at the future time t Expressed as: Where, L i ,W i Indicates the length and width of the obstacle, Indicates the heading and geometric center coordinates of the obstacle; S62: Construct an elliptical region ε of equal probability based on the covariance matrix of the prediction results t , represents the possible location area of obstacles within the confidence range, and the center of the elliptical area is The major axis w and minor axis h of the elliptical area are determined by the eigenvalues λ1 and λ2 of the covariance matrix: Where α is the scaling factor; S63: Quantify the uncertainty occupied area of the obstacle vehicle at time t The risk brought by using the covariance matrix in the forecast information Quantify the uncertainty risk of the occupied area. For the trajectory point in the trajectory τ, the uncertainty risk value at time t is Where, represents the uncertainty risk quantitative index of the occupied area at time t, σ xy Represents the covariance of the prediction results in the x and y directions, σ x Indicates the variance of the prediction result in the x direction, σ y Indicates the variance of the prediction result in the y direction; S64: By correlating the spatiotemporal elements of each trajectory point on the trajectory τ and the predicted occupied area, the uncertainty risk of each point on the trajectory is accumulated to obtain the uncertainty risk of the trajectory 8. The trajectory generation and risk assessment method for urban autonomous driving according to claim 7, characterized in that: In step S7, the comprehensive risk of trajectory τ for: Where η dyn ,η det ,η unc They represent the weight coefficients of the vehicle's dynamic risk, deterministic risk, and uncertainty risk respectively.

9. The trajectory generation and risk assessment method for urban autonomous driving according to claim 8, characterized in that: Step S8 is based on the comprehensive risk of each trajectory obtained in S7. Select Minimum trajectory τ * Output as the optimal trajectory.

Images