Vehicle control method and device, electronic equipment and medium

Abstract

This application provides a vehicle control method, device, electronic device, and medium, relating to the field of autonomous driving technology. It constructs a three-part loss function comprising decision variables, state variables, and control variables, and sums the outputs of these three components at each time step to obtain a first sum. The first sums from multiple time steps are then accumulated to form a unified target sum. Under the constraints of a first vehicle's dynamic mathematical model and first constraints, a sequence of driving variables that minimizes this target sum is jointly solved. Finally, the vehicle is directly controlled using this optimal target driving variable sequence. This effectively bridges the coupling channel between discrete decision-making (i.e., lane decision) and continuous trajectory planning, enabling lane decision-making and trajectory generation to work synchronously and collaboratively under the same optimization objective. This improves the safety, driving efficiency, and ride comfort of autonomous vehicles in complex traffic scenarios, achieving superior overall driving performance compared to traditional fragmented algorithm designs.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and more specifically to a vehicle control method, device, electronic device, and medium. Background Technology

[0002] In existing autonomous driving systems, lane decision-making (e.g., "whether to change lanes, and to which lane") and trajectory planning (e.g., "what curve to take, how much throttle to apply, and how to steer") are typically designed separately. That is, the algorithm first determines whether to change lanes and to which lane, and then performs trajectory planning based on this decision. This fragmented algorithm design results in poor coordination between lane decision-making and trajectory planning. Lane decision-making only aims for "safe lane changes without collisions," while trajectory planning only aims for "the smoothest or most time-efficient" path. The two goals are not aligned, easily leading to problems such as safe decisions but uneven trajectories, or smooth trajectories but unsafe decisions, thus affecting the overall driving performance of autonomous vehicles. Summary of the Invention

[0003] The purpose of this application is to provide a vehicle control method, device, electronic device, and medium, which aims to coordinate lane decision-making and trajectory planning during autonomous driving to enhance the consistency between lane decision-making results and trajectory planning results.

[0004] In a first aspect, embodiments of this application provide a vehicle control method, the method comprising:

[0005] Construct the first loss function, the second loss function, and the third loss function;

[0006] The inputs to the first loss function, the second loss function, and the third loss function all include the driving variables of the first vehicle within a single time step. The driving variables include decision variables, state variables, and control variables. The decision variables are used to characterize the target driving lane of the first vehicle. The state variables are used to characterize the driving state and position of the first vehicle. The control variables are used to characterize the control parameters of the first vehicle. The control parameters are used to instruct the first vehicle to change the driving state. The first loss function is associated with the position deviation, the second loss function is associated with the speed deviation, and the third loss function is associated with the control variables. The position deviation is the deviation between the position of the first vehicle and the target driving lane, and the speed deviation is the deviation between the speed of the first vehicle and the reference speed of the target driving lane.

[0007] The outputs of the first loss function, the second loss function, and the third loss function are added together to obtain the first sum.

[0008] The target sum is obtained by summing the first sums over the multiple time steps.

[0009] Based on the dynamic mathematical model of the first vehicle, the driving variables of the multiple time steps that minimize the target sum value are taken as the target driving variable sequence, and the target driving variable sequence satisfies the first constraint condition.

[0010] The first vehicle is controlled to drive within multiple time steps based on the target driving variable sequence.

[0011] In some implementations, determining the driving variables at multiple time steps that minimize the target sum value, based on the dynamic mathematical model of the first vehicle, as a target driving variable sequence, includes:

[0012] The dynamic mathematical model is linearized to obtain a linearized model;

[0013] The collision cost is added to the target value, wherein the collision cost is associated with a first speed difference and a second speed difference. The first speed difference is the difference between the speed of the target vehicle and the speed of the first vehicle. The target vehicle is a vehicle located in the target driving lane and positioned ahead of the first vehicle in the direction of travel of the first vehicle. The second speed difference is the difference between the speed of a neighboring vehicle and the speed of the first vehicle. The neighboring vehicle is a vehicle whose distance from the first vehicle is less than a preset distance threshold.

[0014] Obtain the driving status information of surrounding vehicles, including the target vehicle and the neighboring vehicles;

[0015] Based on the linearized model and the driving status information of the surrounding vehicles, the driving variables of the multiple time steps that minimize the target sum value are taken as the first driving variable sequence. The first driving variable sequence satisfies the second constraint condition, which includes the first constraint condition.

[0016] Multiple decision variables in the first driving variable sequence are determined as the first variable sequence;

[0017] Using the first variable sequence, a second variable sequence is determined based on the dynamic mathematical model. The second variable sequence includes multiple state variables and multiple control variables for the time steps.

[0018] The first variable sequence is combined with the second variable sequence to obtain the target driving variable sequence.

[0019] In some implementations, determining the second variable sequence based on the dynamic mathematical model using the first variable sequence includes:

[0020] The reference position and reference speed of the vehicle in multiple time steps are determined based on the first variable sequence, and the reference position and reference speed of the vehicle in multiple time steps are used as a reference state variable sequence.

[0021] The target loss value is determined based on the deviation between the vehicle's state variables and the reference state variable sequence in multiple time steps, and the vehicle's control variables in multiple time steps.

[0022] Based on the aforementioned dynamic mathematical model, the state variables and control variables of the multiple time steps that minimize the target loss value are taken as the second variable sequence. The second variable sequence satisfies the third constraint condition, which includes the first constraint condition.

[0023] In some implementations, the third constraint includes the first vehicle being located outside the target ellipse of each of the surrounding vehicles, wherein the target ellipse is an ellipse centered on the position of the surrounding vehicles, and the target ellipse is determined based on the preset major axis and preset minor axis of the surrounding vehicles, and the heading angle of the surrounding vehicles.

[0024] In some implementations, the control variables include the acceleration and steering angle of the first vehicle, and the third constraint condition includes the acceleration of the first vehicle being greater than or equal to a first preset acceleration and less than or equal to a second preset acceleration, and the steering angle of the first vehicle being greater than or equal to a first preset steering angle and less than or equal to a second preset steering angle.

[0025] In some embodiments, the second constraint includes a first difference between the speed of the target vehicle and the speed of the first vehicle being greater than or equal to the negative of a first product of a first preset parameter and a preset binary variable; the second constraint also includes a second difference between the first difference and a second preset parameter, wherein the second product is the product of the first preset parameter and a third difference, and the third difference is the difference between 1 and the preset binary variable, wherein the preset binary variable is 0 or 1.

[0026] In some implementations, the control variables include the acceleration and steering angle of the first vehicle, and the output of the third loss function is the sum of a third product and a fourth product. The third product is the product of a first preset weight and the square of the steering angle of the first vehicle, and the fourth product is the product of a second preset weight and the square of the acceleration of the first vehicle.

[0027] Secondly, embodiments of this application provide a vehicle control device, the device comprising:

[0028] The building module is used to construct the first loss function, the second loss function, and the third loss function;

[0029] The inputs to the first loss function, the second loss function, and the third loss function all include the driving variables of the first vehicle within a single time step. The driving variables include decision variables, state variables, and control variables. The decision variables are used to characterize the target driving lane of the first vehicle. The state variables are used to characterize the driving state and position of the first vehicle. The control variables are used to characterize the control parameters of the first vehicle. The control parameters are used to instruct the first vehicle to change the driving state. The first loss function is associated with the position deviation, the second loss function is associated with the speed deviation, and the third loss function is associated with the control variables. The position deviation is the deviation between the position of the first vehicle and the target driving lane, and the speed deviation is the deviation between the speed of the first vehicle and the reference speed of the target driving lane.

[0030] The first calculation module is used to add the outputs of the first loss function, the second loss function, and the third loss function to obtain a first sum value;

[0031] The second calculation module is used to sum the first sum values ​​of the multiple time steps to obtain the target sum value;

[0032] The determination module is used to determine, based on the dynamic mathematical model of the first vehicle, the driving variables of the multiple time steps that minimize the target sum value as the target driving variable sequence, the target driving variable sequence satisfying the first constraint condition;

[0033] The control module is used to control the first vehicle to drive within multiple time steps based on the target driving variable sequence.

[0034] Thirdly, embodiments of this application provide an electronic device, including:

[0035] The memory is configured to store instructions; and

[0036] The processor is configured to retrieve the instructions from the memory and, when executing the instructions, to implement the vehicle control method provided in the first aspect of the embodiments of this application.

[0037] Fourthly, embodiments of this application provide a machine-readable storage medium storing instructions that, when executed by a processor, cause the processor to implement the vehicle control method according to the first aspect of embodiments of this application.

[0038] In this embodiment, the processor constructs a three-part loss function comprising decision variables, state variables, and control variables: a first loss function quantifies the deviation between the current position of the first vehicle and the center of the target lane; a second loss function quantifies the deviation between the speed of the first vehicle and the reference speed of the target lane; and a third loss function quantifies the smoothness of control variables (such as acceleration and steering angle). The outputs of these three components are summed at each time step to obtain a first sum. The first sums from multiple time steps are then accumulated to form a unified target sum. Under the constraints of the first vehicle's dynamic mathematical model and the first constraint condition, the processor jointly solves for the sequence of driving variables that minimizes this target sum. Finally, the vehicle is directly controlled using this optimal target sequence of driving variables. This effectively bridges the gap between discrete decision-making (i.e., lane decision) and continuous trajectory planning, enabling lane decision-making and trajectory generation to work synchronously and collaboratively under the same optimization objective. This improves the safety, driving efficiency, and ride comfort of autonomous vehicles in complex traffic scenarios, achieving superior overall driving performance compared to traditional fragmented algorithm designs. Attached Figure Description

[0039] Figure 1 This is a schematic flowchart of the vehicle control method provided in the embodiments of this application;

[0040] Figure 2 This is another schematic flowchart of the vehicle control method provided in the embodiments of this application;

[0041] Figure 3 This is a schematic diagram of the vehicle control device provided in the embodiments of this application;

[0042] Figure 4 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0043] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0044] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0045] Autonomous driving effectively improves driving safety and efficiency in modern intelligent transportation systems. The planning task of autonomous driving can be divided into three parts: route planning, decision making, and trajectory planning. Route planning outputs a high-level path based on the road network, while decision making (i.e., lane decision-making) and trajectory planning focus on lane-level planning. Safety is the primary consideration for autonomous vehicles in all tasks. Furthermore, the decision-making ability of autonomous vehicles is significantly affected by interactions with other traffic participants (including passengers and other vehicles). This consistency requirement emphasizes the necessity of adopting a collaborative approach across all tasks of autonomous driving. The main responsibility of autonomous driving planning tasks is to make decisions (such as lane keeping, lane changing, and overtaking) and generate more specific vehicle movements within the lane. Therefore, the integration of decision making and trajectory planning modules has received increasing attention, motivated by the desire to enhance the consistency of results from decision making and trajectory planning tasks, thereby significantly improving the overall driving performance of autonomous vehicles.

[0046] In existing technologies, lane decision-making and trajectory planning are two core steps in autonomous driving tasks. Although significant progress has been made in the development of algorithms for these two tasks, in existing autonomous driving systems, due to the separation of decision-making and trajectory planning algorithm design, each step may have its own cost function, leading to inconsistent goals between the two. This can easily result in problems such as safe decision-making but uneven trajectory, or smooth trajectory but unsafe decision-making, affecting the overall driving performance of autonomous vehicles.

[0047] Based on this, embodiments of this application provide a vehicle control method, device, electronic device, and medium, which aim to coordinate lane decision-making and trajectory planning during autonomous driving to enhance the consistency between lane decision-making results and trajectory planning results.

[0048] Figure 1 This is a schematic flowchart of the vehicle control method provided in the embodiments of this application, as shown below. Figure 1 As shown, the first aspect of this application provides a vehicle control method, including the following steps S100-S500.

[0049] Step S100: Construct the first loss function, the second loss function, and the third loss function;

[0050] The inputs to the first, second, and third loss functions all include the driving variables of the first vehicle within a single time step. The driving variables include decision variables, state variables, and control variables. The decision variables are used to characterize the target driving lane of the first vehicle, the state variables are used to characterize the driving state and position of the first vehicle, and the control variables are used to characterize the control parameters of the first vehicle. The control parameters are used to instruct the first vehicle to change its driving state. The first loss function is associated with the position deviation, the second loss function is associated with the speed deviation, and the third loss function is associated with the control variables. The position deviation is the deviation between the position of the first vehicle and the target driving lane, and the speed deviation is the deviation between the speed of the first vehicle and the reference speed of the target driving lane.

[0051] Those skilled in the art will understand that the processor executing the vehicle control method provided in the embodiments of this application can be the processor of a vehicle autonomous driving system, which can be the vehicle's MCU (Microcontroller Unit) or ECU (Electronic Control Unit).

[0052] In this step, the processor of the first vehicle can define three loss functions based on the settings and inputs of the technicians. These functions correspond to the three optimization objectives of "lane keeping," "reference speed tracking," and "driving comfort," respectively. Specifically, the first loss function quantifies the deviation between the position of the first vehicle (i.e., the vehicle itself) and the target driving lane within a single time step (which can be quantified using the position of the centerline of the target driving lane). Its inputs are the decision variables (representing left lane change, lane keeping, or right lane change), state variables (including vehicle position, heading, speed, etc.), and control variables (such as acceleration, steering angle, etc.) at the current moment. The first loss function can be expressed by the following formula:

[0053]

[0054] in, Let x(τ) be the output of the first loss function, i.e., the instantaneous cost or instantaneous loss at the τ-th time step, where x(τ) is the state variable, b(τ) is the decision variable, μ(τ) is the control variable, τ is the time step number, and b(τ) satisfies the following formula:

[0055]

[0056] Formula (2) can be understood as:

[0057] At the τ-th time step, there are three mutually exclusive discrete action options:

[0058] α = -1 indicates "left lane change", and the target lane is the lane to the left of the current lane (left lane);

[0059] α = 0 means "stay in the current lane", and the target lane is the current lane;

[0060] α = 1 indicates "right lane change", and the target lane is the lane to the right of the current lane (right lane);

[0061] Using three binary variables b -1 b0 and b1 correspond to these three actions respectively:

[0062] If b -1 If (τ) = 1, then the "left lane change" is executed at the τ-th time step;

[0063] If b0(τ) = 1, then "lane keeping" is executed at the τ-th time step;

[0064] If b1(τ) = 1, then a right lane change is performed at the τ-th time step;

[0065] Constraint ∑ α b α (τ) = 1, which means that each time step is exactly the same and at most one action can be selected.

[0066] Thus, formula (1) can be further understood as:

[0067] It represents "the deviation between the position of the first vehicle and the center line of the left lane if the vehicle changes lanes to the left";

[0068] It represents "the deviation between the position of the first vehicle and the centerline of the current lane if the current lane is maintained";

[0069] This characterizes the deviation between the position of the first vehicle and the center line of the right lane if the vehicle changes lanes to the right.

[0070] Because of b α With only one term being 1, Formula (1) aims to identify the distance deviation corresponding to the lane decision and use it as the output of the first loss function, representing the loss for that part.

[0071] The second loss function measures the difference between the current longitudinal speed of the first vehicle and the reference speed of the selected target lane. The input also includes the three types of driving variables mentioned above, thus measuring the speed tracking performance. The second loss function can be expressed by the following formula:

[0072]

[0073] in, The output of the second loss function is the instantaneous cost or instantaneous loss at the τ-th time step. It represents the deviation between the speed of the first vehicle and the reference speed of the left lane if the vehicle changes lanes to the left. It represents "the deviation between the speed of the first vehicle and the reference speed of the current lane if it remains in the current lane". It represents the deviation between the position of the first vehicle and the reference speed of the right lane if the vehicle changes lanes to the right.

[0074] The third loss function, "used to evaluate the smoothness of the control input," is specifically determined by the control variables and reflects the impact on passenger comfort during acceleration and steering. The third loss function can be expressed by the following formula:

[0075]

[0076] in, Let δ(τ) be the output of the third loss function, i.e., the instantaneous cost or instantaneous loss at the τ-th time step, δ(τ) be the turning angle at the τ-th time step, a(τ) be the acceleration at the τ-th time step, and ω be the acceleration. δ As the first preset weight corresponding to the steering angle, ω a The second preset weight is the weight corresponding to the acceleration.

[0077] By simultaneously inputting these three types of variables in a single time step, the processor can obtain the respective outputs of the three loss functions at that moment.

[0078] Step S200: Add the outputs of the first loss function, the second loss function, and the third loss function to obtain the first sum.

[0079] After constructing the three loss functions, the second step is to sum their outputs within the same time step to obtain the first sum. This sum accurately describes the costs incurred by the first vehicle in terms of lane departure, speed deviation, and comfort if it executes the current decision variables, state, and control within this time step. By simply adding the three cost functions, we can obtain a unified quantity that can be directly used to compare the merits of different strategies. The first sum can be expressed by the following formula:

[0080]

[0081] in, This is the first sum.

[0082] Step S300: Sum the first sums of multiple time steps to obtain the target sum.

[0083] In this step, the processor does not only focus on the performance at a single instant, but accumulates all the initial sums over the planning time domain—for example, several discrete time steps in the future—to obtain the target sum for the entire planning period. This accumulation process integrates the cost levels of each time step, ensuring that the optimization process considers both current lane keeping and speed tracking, as well as the smoothness and safety of the entire process, thus forming a globally optimal evaluation metric. The target sum can be expressed using the following formula:

[0084]

[0085] Where J is the target sum, and T-1 is the total number of time steps.

[0086] Step S400: Based on the dynamic mathematical model of the first vehicle, the driving variables of multiple time steps that minimize the target sum are taken as the target driving variable sequence, and the target driving variable sequence satisfies the first constraint condition.

[0087] After obtaining the expressions for the objective and its value, in this step, joint optimization is performed under the first constraint, based on the dynamic mathematical model of the first vehicle (which serves as another constraint for optimization). The optimization process searches for a set of decision variables b(τ), state variables x(τ), and control variables μ(τ) that minimizes the objective and its value across the entire time domain. This set of variables, arranged in chronological order, constitutes the objective driving variable sequence. It clearly defines which lane, speed, acceleration, and steering commands the vehicle should take at each future time step, while ensuring that the global optimization objective is satisfied to the maximum extent.

[0088] The driving variables at multiple time steps that minimize the target sum can be expressed by the following formula:

[0089]

[0090] Wherein, the constraints (including the first constraint) following subjectto indicate the process of determining the driving variables at multiple time steps that minimize the target and its value, including:

[0091] x(τ+1)=f(x(τ),μ(τ)): indicates that x(τ+1) (the τ+1 form of the state variable) needs to satisfy the dynamic mathematical model characterized by f(x(τ),μ(τ));

[0092] x(τ)∈X,μ(τ)∈U: This indicates that x(τ) and μ(τ) must be within the sets X and U, respectively.

[0093] g(x(τ))≥0: This indicates that the first vehicle needs to maintain a certain speed difference or safe distance from other surrounding vehicles to prevent collisions, i.e., collision constraint or obstacle avoidance constraint.

[0094] x(0) = x0: indicates that the initial value of the state variable is x0;

[0095] ∑ α b α (τ)=1、b α (τ)∈{0,1}、α∈{-1,0,1}: Indicator of the constraints defined by formula (2);

[0096] τ = 0, 1, ..., T-1: Indicates the value of τ;

[0097] As mentioned above, the constraints here include both the first vehicle dynamics mathematical model (ensuring that the state and control meet physical feasibility), and discrete decision mutual exclusion (only one of left lane change, straight-ahead, or right lane change is allowed at each moment), and state / control boundaries (speed, acceleration, and steering angle are within safe ranges). The first constraint includes:

[0098] x(0) = x0,

[0099] τ=0,1,…,T-1

[0100] The dynamic mathematical model characterized by f(x(τ),μ(τ)) can be expressed by the following formula:

[0101] p x (τ+1)=p x (τ)+f r (ν(τ),δ(τ))cos(θ(τ)) (8)

[0102] p y (τ+1)=p y (τ)+f r (v(τ),δ(τ))sin(θ(τ)) (9)

[0103]

[0104] ν(τ+1)=ν(τ)+τ s a(τ) (11)

[0105] Where, p x p y Let θ and ν be the x and y coordinates of the first vehicle's center of mass in the ground coordinate system, respectively, together representing the position of the first vehicle; let θ be the heading angle of the first vehicle; let ν be the velocity of the first vehicle; let h be the front and rear wheelbase of the first vehicle; and let τ be the distance between the front and rear wheels. s The sampling time interval for vehicle speed.

[0106] Step S500: Control the movement of the first vehicle within multiple time steps based on the target driving variable sequence.

[0107] In this step, the processor progressively sends the solved target driving variable sequence to the first vehicle. At each time step, the processor switches lanes or maintains lane based on the decision variables specified in the sequence, while simultaneously adjusting the throttle, brakes, and steering according to the corresponding control variables to accurately track the optimized state path. By continuously executing decisions, states, and controls in conjunction, the first vehicle can achieve a safe, efficient, and comfortable driving experience on real roads.

[0108] Through steps S100-S500, the processor constructs a three-part loss function comprising decision variables, state variables, and control variables: a first loss function quantifies the deviation between the current position of the first vehicle and the center of the target lane; a second loss function quantifies the deviation between the speed of the first vehicle and the reference speed of the target lane; and a third loss function quantifies the smoothness of control variables (such as acceleration and steering angle). The outputs of these three components are summed at each time step to obtain a first sum. The first sums from multiple time steps are then accumulated to form a unified target sum. Under the constraints of the first vehicle's dynamic mathematical model and the first constraint condition, the processor jointly solves for the sequence of driving variables that minimizes this target sum. Finally, the vehicle is directly controlled using this optimal target sequence of driving variables. This effectively bridges the gap between discrete decision-making (i.e., lane decision) and continuous trajectory planning, enabling lane decision-making and trajectory generation to work synchronously and collaboratively under the same optimization objective. This improves the safety, driving efficiency, and ride comfort of autonomous vehicles in complex traffic scenarios, achieving superior overall driving performance compared to traditional fragmented algorithm designs.

[0109] In some implementations, the control variables include the acceleration and steering angle of the first vehicle, and the output of the third loss function is the sum of the third product and the fourth product. The third product is the product of the first preset weight and the square of the steering angle of the first vehicle, and the fourth product is the product of the second preset weight and the square of the acceleration of the first vehicle.

[0110] As shown in formula (4), in this embodiment, the output of the third loss function is the third product ω. δ (δ(τ)) 2 With the fourth product ω a (a(τ)) 2 The sum of the values, the third product is the first preset weight ω δ The product of the first vehicle's steering angle δ(τ) and the square of the first vehicle's steering angle, the fourth product is the second preset weight ω. a The product of the acceleration a(τ) of the first vehicle and the square of the acceleration.

[0111] In some implementations, based on the dynamic mathematical model of the first vehicle, multiple time steps of driving variables that minimize the target sum are determined as a sequence of target driving variables, including:

[0112] The dynamic mathematical model is linearized to obtain a linearized model;

[0113] The collision cost is added to the target value, which is associated with the first speed difference and the second speed difference. The first speed difference is the difference between the speed of the target vehicle and the speed of the first vehicle. The target vehicle is a vehicle located in the target driving lane and positioned ahead of the first vehicle in the direction of travel of the first vehicle. The second speed difference is the difference between the speed of the adjacent vehicle and the speed of the first vehicle. The adjacent vehicle is a vehicle whose distance from the first vehicle is less than a preset distance threshold.

[0114] Obtain the driving status information of surrounding vehicles, including the target vehicle and neighboring vehicles;

[0115] Based on the linearized model and the driving status information of surrounding vehicles, the driving variables of multiple time steps that minimize the target sum value are taken as the first driving variable sequence. The first driving variable sequence satisfies the second constraint condition, which includes the first constraint condition.

[0116] Multiple decision variables in the first driving variable sequence are defined as the first variable sequence;

[0117] Using the first variable sequence, based on the dynamic mathematical model, the second variable sequence is determined. The second variable sequence includes state variables and control variables at multiple time steps.

[0118] The first variable sequence is combined with the second variable sequence to obtain the target driving variable sequence.

[0119] In this embodiment, a two-stage optimization method is adopted to effectively solve the original nonlinear and nonconvex lane decision-trajectory planning problem, so that joint decision-making and planning can ensure real-time performance while taking into account safety, efficiency and comfort.

[0120] First, in the first stage, the processor constructs a corresponding linearized model for the high-fidelity nonlinear dynamic mathematical model of the first vehicle, using it to approximate the complex nonlinear dynamics. The linearized model can be expressed by the following formula:

[0121]

[0122] in, For state variables in a linearized model, For control variables in a linearized model, For the linearized model, Ad B d The system matrix, obtained by discretizing the high-fidelity dynamic mathematical model, can be calculated from the physical parameters of the first vehicle and the sampling period.

[0123] Thus, model linearization significantly reduces the computational complexity of formula (7), allowing the subsequent optimization problem to be expressed in a linear form. Next, the processor removes the non-convex collision avoidance constraints involving the speed difference (i.e., the first speed difference) between the first vehicle and the target vehicle (located in the target driving lane and positioned ahead of the first vehicle in its direction of travel) and the speed difference (i.e., the second speed difference) between the first vehicle and neighboring vehicles (vehicles whose distance from the first vehicle is less than a preset distance threshold). These constraints are no longer considered hard conditions, but are instead incorporated into the integrated target and value through constraint relaxation, serving as the collision cost of the soft penalty term. The incorporation process can be expressed as the following formula:

[0124]

[0125] It can be seen that the specific implementation of adding the collision cost to the target sum is to add the collision cost corresponding to that time step to the first sum at each time step, where, The first sum, To create value through collision. This is the portion of the collision cost related to the first velocity difference. This is the portion of the collision cost related to the second velocity difference.

[0126] Thus, the nonlinear and nonconvex parts of formula (7) are "approximated and relaxed" into a form that can be handled by mixed integer programming (MIP). The processor then uses the driving state information of the linearized model and surrounding vehicles (neighboring vehicles + target vehicle) as the first sequence of driving variables for multiple time steps that minimize the target sum. This process is carried out under the second constraint and can be expressed as the following formula:

[0127]

[0128] After obtaining the MIP problem as shown in formula (14) above, the processor can use the branch and bound algorithm to solve it. This algorithm continuously branches the solution space—dividing the possible decisions into several subsets; and delimits each subset (calculating the lower and upper bounds) to gradually approach the optimal solution, thus ensuring both the quality of the solution and the efficiency of online computation.

[0129] After solving the first-stage MIP problem as described in formula (14) above, the processor not only obtains the optimal decision variable sequence b(τ), but also simultaneously calculates the linear state variables at the corresponding time step. and linear control variables This forms a first sequence of driving variables. This set of solutions is the initial solution to the original nonlinear, nonconvex problem, providing a guiding starting point for subsequent large-scale nonlinear trajectory optimization and improving the convergence speed and accuracy of the subsequent second-stage solution.

[0130] In the second stage, based on this initial solution, the processor extracts decision variables from multiple time steps in the first driving variable sequence as the first variable sequence, and reintroduces a high-fidelity nonlinear dynamic mathematical model and strict collision avoidance constraints (g(x(τ))≥0). The processor then uses the first variable sequence to perform refined trajectory optimization to obtain the second variable sequence (including state variables and control variables from multiple time steps, satisfying the third constraint condition) to determine the final output of a smooth, feasible, and safety-constrained complete target driving variable sequence.

[0131] In this way, the processor uses four techniques—model linearization, relaxation of collision constraints into soft penalties, branch and bound solution of MIP, and use of the first-stage solution as the initial solution—to successfully simplify the originally difficult mixed-integer nonlinear nonconvex optimization problem into an efficient MIP problem. It also achieves the dual goals of real-time optimization decision-making and high-quality trajectory planning through two-stage collaboration.

[0132] In some implementations, the second constraint includes a first difference between the speed of the target vehicle and the speed of the first vehicle being greater than or equal to the negative of a first product of a first preset parameter and a preset binary variable; the second constraint also includes a first difference being less than or equal to a second difference between a second product and a second preset parameter, the second product being the product of the first preset parameter and a third difference, the third difference being the difference between 1 and a preset binary variable, and the preset binary variable being 0 or 1.

[0133] In this embodiment, in order to incorporate the originally non-convex collision avoidance constraint of "collision risk caused by the speed difference with the target vehicle" into the second constraint condition and maintain linear solvability, the processor determines the following two inequality formulas for the first vehicle speed difference (i.e., the first difference, the first difference between the speed of the target vehicle and the speed of the first vehicle) at the τ-th time step:

[0134]

[0135] in, v is the speed of the target vehicle. x(τ) represents the velocity of the first vehicle (the velocity component in the vehicle's forward direction, hereinafter referred to as the velocity of the first vehicle), and M is a first preset parameter. As a pre-defined binary variable, This is the first difference;

[0136] That is, when a binary variable is pre-defined When ξ = 1, the right end is 0, requiring the speed of the first vehicle to not exceed the speed of the target vehicle, strictly avoiding a rear-end collision; when ξ = 1, the right end is -M, at which point the restrictions are relaxed. The lower bound constraint allows for the temporary "opening" of rear-end collision risk in the planning process, and the value of M is a very large value;

[0137]

[0138] in, The second product is represented by ∈, which is the second preset parameter. The second difference, The third difference is σ+α, where σ is the target lane number, σ is the current lane number, and α indicates the lane change decision (-1, 0, 1).

[0139] That is, when When the right end is -ε, it is required that Δν ≤ -ε, that is, the speed of the first vehicle is higher than the speed of the target vehicle and exceeds a threshold (the value of ε can be a very small value); when When the right end is M-ε, the constraint fails, thus ensuring that there is no upper limit when there is no risk.

[0140] Through formulas (15) and (16), the processor cleverly incorporates the statement "If there is a risk of rear-end collision due to speed difference, then..." And be punished, otherwise and forced The logic of "" has been completely transformed into "the logic of "". With pre-defined binary variables The linear constraints are incorporated into the collision avoidance conditions, which were originally nonlinear and nonconvex, and are now included as a second constraint that can be solved using mixed integer programming (MIP). This "Big-M soft-switching" approach can... Strictly ensure safe distance (hard constraint) while also being able to Time through The corresponding soft penalty term allows the optimizer to weigh the trade-off between "risk" and "comfort / efficiency," thereby improving the solvability and computational efficiency of collision avoidance constraints in real-time online planning.

[0141] At the same time, for adjacent vehicles, constraints such as those in formulas (15) and (16) can also be incorporated into the above-mentioned second constraint:

[0142]

[0143] in, The speed of the nearby vehicles.

[0144] In summary, formula (14) can be written as:

[0145]

[0146]

[0147] The condition following "subject to" indicates the second constraint, including:

[0148] instruct Satisfies the linearization model;

[0149] v x ≥ρ|v y |,v x,min ≥0: Indicates v x Greater than or equal to 0 and greater than or equal to the product of the preset ρ and the absolute value of the lateral velocity component;

[0150] Formula (15)-Formula (18);

[0151] a x,min ≤a x (τ)≤a x,max a y,min ≤a y (τ)≤a y,max : Indicates the upper and lower limits of the components of the first vehicle's acceleration in the forward and lateral directions;

[0152] ∑ α b α (τ)=1、v α (τ)∈{0,1},α∈{-1,0,1};

[0153] And the first constraint condition.

[0154] In some implementations, the second variable sequence is determined using the first variable sequence based on a dynamic mathematical model, including:

[0155] The reference position and reference speed of the vehicle in multiple time steps are determined based on the first variable sequence, and the reference position and reference speed of the vehicle in multiple time steps are used as the reference state variable sequence.

[0156] The target loss value is determined based on the deviation between the vehicle's state variables and the reference state variable sequence over multiple time steps, as well as the vehicle's control variables over multiple time steps.

[0157] Based on the dynamic mathematical model, the state variables and control variables of multiple time steps that minimize the target loss value are taken as the second variable sequence. The second variable sequence satisfies the third constraint condition, which includes the first constraint condition.

[0158] In this embodiment, the second stage uses a high-fidelity dynamic mathematical model to refine the trajectory of the first vehicle based on the first variable sequence b(τ) obtained in the first stage.

[0159] Specifically, the reference position of the first vehicle in each time step is first derived from the target driving lane and the reference speed of the target driving lane at each time step indicated by b(τ). and reference speed These are arranged into a reference state variable sequence. Subsequently, based on the lateral and longitudinal deviations between the state variables of the first vehicle at multiple time steps and this reference state variable sequence, and the control variables μ(τ) (such as acceleration a(τ) and steering angle δ(τ)) at the corresponding time steps, the loss values ​​of the trajectories corresponding to the state variables and control variables at multiple time steps are calculated using the following formula. The target loss value is obtained by summing the loss values ​​of all time steps.

[0160]

[0161] Where q1, q2, q3, r1, and r2 are all preset weight values.

[0162] The processor incorporates a high-fidelity dynamic mathematical model and precise obstacle avoidance constraints into the third constraint condition. Through optimization, it determines the state—the control sequence—that minimizes the target loss value, as the second variable sequence x(τ), μ(τ). This second variable sequence represents the smooth, executable trajectory that conforms to the high-fidelity dynamic mathematical model. The process of determining the second variable sequence can be expressed by the following formula:

[0163]

[0164] Thus, on the one hand, by using the reference trajectory generated in the first stage of decision-making, the search space in the second stage is effectively reduced, and the convergence speed of trajectory optimization is improved; on the other hand, by directly minimizing the position tracking error, speed tracking error and comfort cost in the complete nonlinear dynamic mathematical model, the final output trajectory not only ensures high-precision lane and speed tracking, but also takes into account the smoothness of acceleration and steering, achieving better driving safety, efficiency and ride comfort than the traditional two-stage separation method.

[0165] In some implementations, the third constraint includes the first vehicle being located outside the target ellipse of each of the surrounding vehicles, wherein the target ellipse is an ellipse centered on the position of the surrounding vehicles, and the target ellipse is determined based on the preset major axis and preset minor axis of the surrounding vehicles, and the heading angle of the surrounding vehicles.

[0166] In this embodiment, the obstacle avoidance constraint in the third constraint is expressed as "the position of the first vehicle (which may be the position of its centroid) must be outside the target ellipse defined by each surrounding vehicle." Specifically, for each surrounding vehicle, its current position... As the center of the ellipse, based on the vehicle's heading angle A i Rotate the ellipse in this direction, and then rotate it around the preset major semi-axis. and short half shaft The dimensions of the ellipse are used as the reference point. This yields a "safe zone," or target ellipse, that closely follows the vehicle's geometry and direction of travel. When the position of the first vehicle is substituted into the quadratic inequality shown in the following formula, if this equation is satisfied, it means that the first vehicle falls outside all these target ellipses and will not overlap with the safe zones of any surrounding vehicles:

[0167]

[0168] The above inequality can be expressed as g(x(τ))≥0 in formula (7) after transformation.

[0169] In some implementations, the control variables include the acceleration and steering angle of the first vehicle, and the third constraint includes the acceleration of the first vehicle being greater than or equal to a first preset acceleration and less than or equal to a second preset acceleration, and the steering angle of the first vehicle being greater than or equal to a first preset steering angle and less than or equal to a second preset steering angle.

[0170] In this embodiment, the third constraint also includes setting upper and lower limits for the acceleration and steering angle of the first vehicle:

[0171] a min ≤a(τ)≤a max (twenty two)

[0172] δ min ≤δ(τ)≤δ max (twenty three)

[0173] Among them, a min For the first preset acceleration, a max For the second preset acceleration, δ min The first preset steering angle, δ max This is the second preset steering angle.

[0174] Combining the two implementation methods described above, formula (20) can be expressed as:

[0175]

[0176] Figure 2 This is another schematic flowchart of the vehicle control method provided in the embodiments of this application, please refer to it as well. Figures 1-2 The technical effects of the vehicle control method provided in this application embodiment can be experimentally verified using the following process:

[0177] In the simulation environment, the equipment used for simulation experiments first initializes the state of the first vehicle (including position, speed, and heading), the state information of surrounding vehicles, and inputs such as road topology and traffic rules. Then, it enters the first stage of optimization—using the linearized model and soft collision penalty through the decision-making module to solve for the optimal sequence of decision variables. Next, based on this sequence of decision variables, the second stage of optimization is initiated. Under a high-fidelity dynamic mathematical model and refined elliptical obstacle avoidance constraints, a second sequence of variables that meets the requirements of dynamic constraints, collision avoidance, and path smoothness is iteratively generated (based on the iterative linear quadratic regulator, iLQR, linear-quadratic optimal control algorithm and alternating direction method of multipliers, ADMM, the alternating direction multiplier method, ultimately outputting a coherent sequence of target driving variables (including continuous states and control variables). This sequence of target driving variables is loaded into a simulation software platform, and the simulated first vehicle travels according to the planned trajectory. The simulation is evaluated using safety (number of collisions, number of emergency brakings), efficiency (driving time, energy consumption), and comfort (acceleration and steering angle change rate). Simulation results show that in multi-lane, dynamic traffic scenarios, the vehicle control method provided in this application can effectively handle mixed constraints while ensuring real-time performance, generating a safe, efficient, and smooth driving trajectory, which is significantly better than traditional fragmented designs.

[0178] In a real-world driving environment, engineers deployed the decision-making and trajectory planning modules on an autonomous driving test vehicle (the first vehicle) equipped with sensors such as LiDAR, cameras, and millimeter-wave radar. After the first vehicle collected real-time data on its surroundings and its own state, its processor, following the first and second stage optimization processes described above, first generated a sequence of decision variables, then output a refined sequence of target driving variables, which was then sent to the vehicle control system. Through road testing, evaluated against the same safety, efficiency, and comfort indicators, the vehicle control method provided in this application embodiment also demonstrated excellent obstacle avoidance capabilities, decision-making and planning coordination, and overall improved driving performance on actual roads, verifying the feasibility and effectiveness of the method in real-world scenarios.

[0179] Figure 3 This is a schematic diagram of the vehicle control device provided in the embodiments of this application, such as... Figure 3 As shown, a second aspect of this application provides a vehicle control device 10, comprising:

[0180] Module 11 is used to construct the first loss function, the second loss function, and the third loss function;

[0181] The inputs to the first loss function, the second loss function, and the third loss function all include the driving variables of the first vehicle within a single time step. The driving variables include decision variables, state variables, and control variables. The decision variables are used to characterize the target driving lane of the first vehicle, the state variables are used to characterize the driving state and position of the first vehicle, and the control variables are used to characterize the control parameters of the first vehicle. The control parameters are used to instruct the first vehicle to change its driving state. The first loss function is associated with the position deviation, the second loss function is associated with the speed deviation, and the third loss function is associated with the control variables. The position deviation is the deviation between the position of the first vehicle and the target driving lane, and the speed deviation is the deviation between the speed of the first vehicle and the reference speed of the target driving lane.

[0182] The first calculation module 12 is used to add the outputs of the first loss function, the second loss function, and the third loss function to obtain a first sum value;

[0183] The second calculation module 13 is used to sum the first sums of multiple time steps to obtain the target sum.

[0184] The determination module 14 is used to determine the driving variables of multiple time steps that minimize the target sum value based on the dynamic mathematical model of the first vehicle as the target driving variable sequence, and the target driving variable sequence satisfies the first constraint condition.

[0185] The control module 15 is used to control the movement of the first vehicle in multiple time steps according to the target driving variable sequence.

[0186] The vehicle control device 10 provided in the second aspect of the present application can implement the various processes implemented in the above method embodiments and achieve the same beneficial effects. To avoid repetition, it will not be described again here.

[0187] Please see Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. A third aspect of this application provides an electronic device 1000, including a processor 1100 and a memory 1200. The memory 1200 stores machine-executable instructions that can be executed by the processor 1100. The processor 1100 can execute the machine-executable instructions to implement the above-mentioned vehicle control method.

[0188] A fourth aspect of this application provides a machine-readable storage medium storing instructions that, when executed by a processor, cause the processor to implement the vehicle control method described above.

[0189] In some embodiments, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the vehicle control method according to the above embodiments.

[0190] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0191] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0192] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0193] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0194] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media (transistor silicon switching devices), such as modulated data signals and carrier waves.

[0195] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0196] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

[0197] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed by the present invention.

Claims

1. A vehicle control method, characterized in that, The method includes: Construct the first loss function, the second loss function, and the third loss function; The inputs to the first loss function, the second loss function, and the third loss function all include the driving variables of the first vehicle within a single time step. The driving variables include decision variables, state variables, and control variables. The decision variables are used to characterize the target driving lane of the first vehicle. The state variables are used to characterize the driving state and position of the first vehicle. The control variables are used to characterize the control parameters of the first vehicle. The control parameters are used to instruct the first vehicle to change the driving state. The first loss function is associated with the position deviation, the second loss function is associated with the speed deviation, and the third loss function is associated with the control variables. The position deviation is the deviation between the position of the first vehicle and the target driving lane, and the speed deviation is the deviation between the speed of the first vehicle and the reference speed of the target driving lane. The outputs of the first loss function, the second loss function, and the third loss function are added together to obtain the first sum. The target sum is obtained by summing the first sum values ​​of the multiple time steps. Based on the dynamic mathematical model of the first vehicle, the driving variables of the multiple time steps that minimize the target sum value are taken as the target driving variable sequence, and the target driving variable sequence satisfies the first constraint condition. The first vehicle is controlled to drive within multiple time steps based on the target driving variable sequence; The dynamic mathematical model based on the first vehicle includes a sequence of driving variables at multiple time steps that minimize the target sum, comprising: The dynamic mathematical model is linearized to obtain a linearized model; The collision cost is added to the target value, wherein the collision cost is associated with a first speed difference and a second speed difference. The first speed difference is the difference between the speed of the target vehicle and the speed of the first vehicle. The target vehicle is a vehicle located in the target driving lane and positioned ahead of the first vehicle in the direction of travel of the first vehicle. The second speed difference is the difference between the speed of a neighboring vehicle and the speed of the first vehicle. The neighboring vehicle is a vehicle whose distance from the first vehicle is less than a preset distance threshold. Obtain the driving status information of surrounding vehicles, including the target vehicle and the neighboring vehicles; Based on the linearized model and the driving status information of the surrounding vehicles, the driving variables of the multiple time steps that minimize the target sum value are taken as the first driving variable sequence. The first driving variable sequence satisfies the second constraint condition, which includes the first constraint condition. Multiple decision variables in the first driving variable sequence are determined as the first variable sequence; Using the first variable sequence, a second variable sequence is determined based on the dynamic mathematical model. The second variable sequence includes multiple state variables and multiple control variables for the time steps. The first variable sequence is combined with the second variable sequence to obtain the target driving variable sequence.

2. The method according to claim 1, characterized in that, The step of determining the second variable sequence based on the first variable sequence and the dynamic mathematical model includes: The reference position and reference speed of the vehicle in multiple time steps are determined based on the first variable sequence, and the reference position and reference speed of the vehicle in multiple time steps are used as a reference state variable sequence. The target loss value is determined based on the deviation between the vehicle's state variables and the reference state variable sequence in multiple time steps, and the vehicle's control variables in multiple time steps. Based on the aforementioned dynamic mathematical model, the state variables and control variables of the multiple time steps that minimize the target loss value are taken as the second variable sequence. The second variable sequence satisfies the third constraint condition, which includes the first constraint condition.

3. The method according to claim 2, characterized in that, The third constraint includes the first vehicle being located outside the target ellipse of each of the surrounding vehicles, wherein the target ellipse is an ellipse centered on the position of the surrounding vehicles, and the target ellipse is determined based on the preset major axis and preset minor axis of the surrounding vehicles, as well as the heading angle of the surrounding vehicles.

4. The method according to claim 2, characterized in that, The control variables include the acceleration and steering angle of the first vehicle, and the third constraint condition includes the acceleration of the first vehicle being greater than or equal to the first preset acceleration and less than or equal to the second preset acceleration, and the steering angle of the first vehicle being greater than or equal to the first preset steering angle and less than or equal to the second preset steering angle.

5. The method according to claim 1, characterized in that, The second constraint includes a first difference between the speed of the target vehicle and the speed of the first vehicle being greater than or equal to the negative of a first product of a first preset parameter and a preset binary variable; the second constraint also includes a second difference between the first difference and a second preset parameter, wherein the second product is the product of the first preset parameter and a third difference, and the third difference is the difference between 1 and the preset binary variable, wherein the preset binary variable is 0 or 1.

6. The method according to claim 1, characterized in that, The control variables include the acceleration and steering angle of the first vehicle. The output of the third loss function is the sum of the third product and the fourth product. The third product is the product of the first preset weight and the square of the steering angle of the first vehicle. The fourth product is the product of the second preset weight and the square of the acceleration of the first vehicle.

7. A vehicle control device, characterized in that, The device includes: The building module is used to construct the first loss function, the second loss function, and the third loss function; The inputs to the first loss function, the second loss function, and the third loss function all include the driving variables of the first vehicle within a single time step. The driving variables include decision variables, state variables, and control variables. The decision variables are used to characterize the target driving lane of the first vehicle. The state variables are used to characterize the driving state and position of the first vehicle. The control variables are used to characterize the control parameters of the first vehicle. The control parameters are used to instruct the first vehicle to change the driving state. The first loss function is associated with the position deviation, the second loss function is associated with the speed deviation, and the third loss function is associated with the control variables. The position deviation is the deviation between the position of the first vehicle and the target driving lane, and the speed deviation is the deviation between the speed of the first vehicle and the reference speed of the target driving lane. The first calculation module is used to add the outputs of the first loss function, the second loss function, and the third loss function to obtain a first sum value; The second calculation module is used to sum the first sum values ​​of the multiple time steps to obtain the target sum value; The determination module is used to determine, based on the dynamic mathematical model of the first vehicle, the driving variables of the multiple time steps that minimize the target sum value as the target driving variable sequence, the target driving variable sequence satisfying the first constraint condition; The control module is used to control the driving of the first vehicle within multiple time steps according to the target driving variable sequence; The determining module is further configured to: The dynamic mathematical model is linearized to obtain a linearized model; The collision cost is added to the target value, wherein the collision cost is associated with a first speed difference and a second speed difference. The first speed difference is the difference between the speed of the target vehicle and the speed of the first vehicle. The target vehicle is a vehicle located in the target driving lane and positioned ahead of the first vehicle in the direction of travel of the first vehicle. The second speed difference is the difference between the speed of a neighboring vehicle and the speed of the first vehicle. The neighboring vehicle is a vehicle whose distance from the first vehicle is less than a preset distance threshold. Obtain the driving status information of surrounding vehicles, including the target vehicle and the neighboring vehicles; Based on the linearized model and the driving status information of the surrounding vehicles, the driving variables of the multiple time steps that minimize the target sum value are taken as the first driving variable sequence. The first driving variable sequence satisfies the second constraint condition, which includes the first constraint condition. Multiple decision variables in the first driving variable sequence are determined as the first variable sequence; Using the first variable sequence, a second variable sequence is determined based on the dynamic mathematical model. The second variable sequence includes multiple state variables and multiple control variables for the time steps. The first variable sequence is combined with the second variable sequence to obtain the target driving variable sequence.

8. An electronic device, characterized in that, It includes a processor and a memory, the memory storing programs or instructions that can run on the processor, the programs or instructions being executed by the processor to implement the vehicle control method as described in any one of claims 1-6.

9. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions that, when executed by a processor, cause the processor to implement the vehicle control method as described in any one of claims 1-6.

Images