Pipeline-based model predictive control trajectory planning method

Abstract

The invention relates to a pipeline-based model predictive control trajectory planning method, and the method comprises the steps: collecting the environment information around a vehicle, and enabling the environment information to comprise the information related to an obstacle; generating a reference pipeline representing a travelable space of the own vehicle at least based on the steering limitation of the own vehicle and the predicted trajectory of the obstacle; when the optimization problem of the model prediction control is solved, the center line of the reference pipeline serves as a reference value of the optimization problem, and therefore the track of the self-vehicle is generated, and the self-vehicle can run along the generated track. The application also relates to a computer program product and an electronic device.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and more particularly, to a trajectory planning method based on pipelines for model predictive control (MPC), a method for adaptive cruise control, a computer program product, and an electronic device. Background Art

[0002] In the field of autonomous driving, path planning is an important link, which involves finding a safe, feasible, and efficient path for a vehicle from a starting point to an ending point in a given environment. In a dynamic environment, the vehicle needs to perceive and respond to changes in the surrounding environment in real time. Existing path planning methods (such as model predictive control algorithms) may not be able to solve the optimization problem within a limited time, resulting in untimely path updates and affecting the real-time performance of the vehicle. That is to say, existing path planning methods have certain limitations, such as too high computational complexity, and often cannot balance accuracy and efficiency well. Summary of the Invention

[0003] The inventors of the present application have realized that in autonomous driving path planning, the solution time of the optimization problem is a key factor. Traditional methods have a large amount of calculation when dealing with complex environments, resulting in a long solution time and being difficult to meet the real-time requirements. In addition, traditional methods are usually based on kinematic models, which, although simple to calculate, cannot fully capture the dynamic behavior of the vehicle, resulting in the generated path being possibly infeasible during actual execution.

[0004] To solve this problem, the inventors of the present application propose to gradually narrow the search space of the optimization problem by constructing a feasible region (such as a hard constraint pipeline, a soft constraint pipeline, and a reference pipeline), thereby improving the solution efficiency. This trajectory planning method can not only ensure the safety of the vehicle during driving, but also quickly generate a path that meets the vehicle dynamics constraints and takes into account the influence of the predicted trajectories of obstacles in a complex dynamic environment. By reasonably constructing and narrowing the feasible region, the solution time of the optimization problem can be significantly reduced on the premise of ensuring the safety and feasibility of the path, and the real-time performance and practicality of the entire autonomous driving system can be improved.

[0005] According to one aspect of the present application, the present application provides a trajectory planning method for model predictive control (MPC) based on a pipeline, the method comprising: collecting environmental information around the ego vehicle, the environmental information including information related to obstacles; generating a reference pipeline representing the drivable space of the ego vehicle based at least on the steering limit of the ego vehicle and the predicted trajectories of the obstacles; and when solving the optimization problem of the model predictive control, using the center line of the reference pipeline as the reference value of the optimization problem, so as to generate the trajectory of the ego vehicle such that the ego vehicle can drive along the generated trajectory.

[0006] As a supplement or replacement to the above solution, in the above method, generating a reference pipeline representing the drivable space of the ego vehicle by combining the steering limit of the ego vehicle and the predicted trajectories of the obstacles includes: scanning along a reference line to generate a hard constraint pipeline, the hard constraint pipeline representing a first path area where the ego vehicle can safely pass; on the basis of the hard constraint pipeline, generating a soft constraint pipeline by adding lateral and longitudinal buffer zones, the soft constraint pipeline representing a second path area where the ego vehicle can safely pass, and the second path area being smaller than the first path area; and generating a reference pipeline on the basis of the soft constraint pipeline by considering the influence of the steering limit of the ego vehicle and the predicted trajectories of the obstacles, the reference pipeline representing a third path area where the ego vehicle can safely pass, and the third path area being smaller than the second path area.

[0007] As a supplement or replacement to the above solution, in the above method, the environmental information further includes lane boundaries sensed by sensors, and the reference line is generated depending on the lane boundaries.

[0008] As a supplement or replacement to the above solution, in the above method, the predicted trajectories of the obstacles are determined according to the predicted trajectories of the obstacles.

[0009] As a supplement or replacement to the above solution, in the above method, the optimization problem of the model predictive control is to minimize an objective function to find the optimal control input such that the planned trajectory of the ego vehicle is as close as possible to the reference value while satisfying the constraint conditions.

[0010] As a supplement or replacement to the above solution, in the above method, the objective function is composed of a position error term, a steering rate change term, an environmental constraint term, and an acceleration constraint term, and the constraint conditions include dynamic constraints, steering actuator limits, acceleration constraints, and environmental constraints.

[0011] As a supplement or replacement to the above solution, in the above method, the environmental information is received from multiple sensors, and the method further includes providing the generated trajectory of the ego vehicle to lateral and longitudinal actuators for controlling the ego vehicle.

[0012] According to another aspect of the present application, there is provided a computer program product including a computer program which, when executed by a processor, implements the method described above.

[0013] According to yet another aspect of the present application, there is provided an electronic device (such as a domain controller, a camera or a radar), including a memory, a processor, and a computer program stored on the memory, where the processor executes the computer program to implement the method described above.

[0014] In the pipeline-based model predictive control trajectory planning scheme of the embodiments of the present application, the vehicle dynamics constraints and the influence of the predicted trajectories of obstacles are considered during the construction of the pipeline (i.e., tube, such as the reference pipeline), so as to ensure that the vehicle will not collide with obstacles during driving while narrowing the feasible region of the optimization problem, thereby enabling the rapid generation of a collision-free trajectory that conforms to the dynamics constraints.

[0015] In one or more embodiments, the solution of the present application is based on a dynamics model, which better meets the actual control requirements (control feasible). The dynamics model takes into account the dynamic characteristics of the vehicle such as inertia, acceleration, and steering rate, and can more accurately describe the motion behavior of the vehicle.

[0016] In one or more embodiments, by generating a hard constraint pipeline, a soft constraint pipeline, and a reference pipeline, the feasible region of the optimization problem is gradually narrowed, and the search space of the optimization problem is reduced, thereby improving the solution efficiency. A smaller feasible region means a lower complexity of the optimization problem, and the solver can more efficiently find the optimal path that satisfies all the constraint conditions, meeting the real-time requirements.

[0017] In one or more embodiments, the objective function of the optimization problem incorporates position error, steering rate change, environmental constraints, acceleration constraints, etc. Different objectives are balanced through reasonable weight coefficients to achieve multi-objective optimization. For example, by optimizing the steering rate change and acceleration change, the frequent steering of the vehicle during driving is reduced, enhancing the comfort experience of passengers. Additionally, slack variables are introduced when solving the optimization problem to allow slight violation of the constraints in infeasible cases, ensuring that the optimization problem can find a feasible solution.

[0018] The model predictive control algorithm only considers the optimization problem within a finite time range at each time step, which may cause the vehicle to fall into a local optimal solution rather than a global optimal solution. In one or more embodiments, the pipeline-based model predictive control (MPC) trajectory planning scheme of the embodiments of the present application relies on the reference line provided by the global path planning to generate the hard constraint pipeline, the soft constraint pipeline, and the reference pipeline, so as to obtain the global optimal solution. Description of the Drawings

[0019] The above and other objects and advantages of the present application will become more fully apparent from the following detailed description taken in conjunction with the accompanying drawings, in which like or similar elements are denoted by like reference numerals.

[0020] Figure 1 FIG. shows a schematic flow chart of a pipeline-based model predictive control trajectory planning method according to an embodiment of the present application;

[0021] Figure 2 FIG. shows a schematic diagram of a hard constraint pipeline, a soft constraint pipeline, and a reference pipeline according to an embodiment of the present application; and

[0022] Figure 3 FIG. shows a schematic structural diagram of an electronic device according to an embodiment of the present application. Detailed Embodiments

[0023] Hereinafter, a pipeline-based model predictive control trajectory planning scheme according to various exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0024] In the description of this specification, the description of reference terms such as "an embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the embodiments of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance. In addition, without contradiction, those skilled in the art can combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples.

[0025] Figure 1 FIG. shows a schematic flow chart of a pipeline-based model predictive control trajectory planning method 1000 according to an embodiment of the present application. As Figure 1 shown, the method 1000 includes the following steps:

[0026] In step S110, environmental information around the host vehicle is collected, and the environmental information includes information related to obstacles;

[0027] In step S120, at least based on the steering limit of the host vehicle and the predicted trajectory of the obstacle, a reference pipeline representing the drivable space of the host vehicle is generated; and

[0028] In step S130, when solving the optimization problem of the model predictive control, the center line of the reference pipeline is used as the reference value of the optimization problem, so as to generate the trajectory of the host vehicle such that the host vehicle can travel along the generated trajectory.

[0029] Model Predictive Control (MPC) is an advanced control strategy that determines the optimal control input by solving an optimization problem with a finite time horizon at each time step. The core idea of MPC is that at each sampling moment, based on the current state and the model, it predicts the system behavior in the future for a period of time, and finds the control sequence that optimizes the performance index through an optimization algorithm.

[0030] In the context of this application, the MPC trajectory planning method 1000 based on the pipeline narrows the feasible region of the optimization problem by constructing a reference pipeline, thereby improving the solution efficiency. This method 1000 combines the advantages of global path planning and local path planning, ensuring that the vehicle can maintain a safe distance and conform to the dynamic characteristics of the vehicle during driving.

[0031] In step S110, the environmental information around the host vehicle is collected. In the context of this application, the term "host vehicle" refers to a vehicle that is in motion and equipped with relevant sensors and control systems. For example, in autonomous driving technology, the host vehicle senses the surrounding environment through various sensors and makes corresponding driving decisions and control actions based on the sensed information. In one embodiment, the environmental information includes information related to obstacles, such as position and speed (used to determine the position and motion speed of the obstacle in the world coordinate system and predict its future position), size and shape (used to understand the size and shape of the obstacle and evaluate the spatial relationship between the vehicle and the obstacle), and category (used to identify the type of the obstacle, such as pedestrians, bicycles, cars, etc., and different categories may require different coping strategies), etc.

[0032] In one or more embodiments, the environmental information around the host vehicle can be obtained through sensor data. The sensors may include (but are not limited to) LiDAR, cameras, and radars. For example, LiDAR can obtain the three-dimensional point cloud data of the surrounding environment by emitting laser beams and measuring the reflection time, and accurately identify the position and shape of obstacles. Cameras can capture visual images and identify obstacles, lane lines, traffic signs, and traffic lights, etc. through image processing and computer vision techniques. Radars can obtain the distance and relative speed of obstacles by emitting radio waves and measuring the reflection time.

[0033] In step S120, a reference tube representing the drivable space of the host vehicle is generated based at least on the steering limit of the host vehicle and the predicted trajectory of the obstacle. Here, the "steering limit" refers to the capacity limit of the vehicle's steering system during driving. For example, the steering limit may include the following aspects:

[0034] (1) Maximum steering angle: the maximum turning angle of the vehicle's steering wheel or steering column. Exceeding this angle may cause the vehicle to lose control or damage the steering system; and

[0035] (2) Steering rate limit: the maximum rate of change of the vehicle's steering angle. This limits the speed at which the vehicle can change direction per unit time and affects the vehicle's maneuverability.

[0036] In other embodiments, the steering limit may further include the minimum turning radius (i.e., the minimum radius at which the vehicle can turn) and the steering response time (i.e., the time required for the vehicle to actually start steering from the moment it receives a steering command), etc.

[0037] The "predicted trajectory of the obstacle" refers to the possible movement path of surrounding obstacles (such as other vehicles, pedestrians, etc.) in the future for a period of time by analyzing and / or modeling their motion states. This prediction is crucial for generating a safe and feasible vehicle trajectory because it allows the autonomous driving system to plan in advance to avoid collisions with obstacles. In one embodiment, the predicted trajectory of the obstacle can be determined based on information such as its current position and speed. By considering the motion speed and direction of the obstacle, its future position is predicted, thereby evaluating its potential threat to the host lane.

[0038] In the context of this application, the "reference tube" defines the space through which the vehicle can safely pass during driving. It takes into account factors such as the vehicle's size, dynamic characteristics, and obstacles in the surrounding environment, ensuring that the vehicle can maintain a safe distance during driving and conform to the vehicle's dynamic characteristics.

[0039] In one embodiment, the reference tube can be generated in the following manner:

[0040] (1) Generate a hard constraint tube

[0041] The hard constraint tube is the first path area generated by scanning along the reference line, which represents the initial path area where the vehicle can safely pass. In one embodiment, generating the hard constraint tube by scanning along the reference line may include: setting the starting point and ending point of the scan, for example, starting from the current position of the vehicle to the target position; gradually evaluating the distance between the vehicle and the road boundary along the reference line to ensure that these distances are greater than the safety distance threshold of the vehicle; and determining the boundary points of the hard constraint tube based on the evaluated distances, and these boundary points define the area where the vehicle can safely pass. In one embodiment, the hard constraint tube is defined by the physical road boundary. For example, the width and shape of the hard constraint tube can be determined according to the size of the vehicle, the safety margin, and the road boundary information. The width of the hard constraint tube is usually set according to the width of the vehicle and the safety distance to ensure that the vehicle does not exceed the road boundary when driving within the hard constraint tube. In this way, the hard constraint tube provides the most basic feasible path for the vehicle to ensure that the vehicle does not exceed the road boundary during driving.

[0042] (2) Generate the soft constraint tube

[0043] The soft constraint tube is the second path area generated by adding horizontal and vertical buffers on the basis of the hard constraint tube, which represents the extended path area where the vehicle can safely pass. In one embodiment, the soft constraint tube can be generated by adding a buffer zone with a certain width on the basis of the hard constraint tube, which can provide an additional safety margin. The width of the buffer zone can be dynamically adjusted according to the speed of the vehicle, its dynamic characteristics, and the environmental complexity. The soft constraint tube adds an additional safety margin to ensure that the vehicle has enough space to avoid obstacles and adjust the path in a complex environment.

[0044] (3) Generate the reference tube

[0045] The reference tube is the final path area generated by combining the steering limit of the vehicle and the predicted trajectory of the obstacle on the basis of the soft constraint tube, which represents the final path area where the vehicle can safely pass. In one embodiment, the reference tube can be generated by further refining and adjusting the soft constraint tube by combining the steering limit of the vehicle (such as the maximum steering angle, the minimum turning radius) and the predicted trajectory of the obstacle (such as the minimum distance between the obstacle and the vehicle, the predicted trajectory of the obstacle). The width and shape of the reference tube will change dynamically according to these factors to ensure that the vehicle can maintain a safe distance and conform to the dynamic characteristics of the vehicle during driving. In this embodiment, the reference tube provides the final feasible domain to ensure that the vehicle can maintain a safe distance and conform to the dynamic characteristics of the vehicle during driving.

[0046] Reference Figure 2, which shows a schematic diagram of a hard-constraint pipeline, a soft-constraint pipeline, and a reference pipeline according to an embodiment of the present application. As Figure 2 shown, the host vehicle 210 travels in a lane formed by lane lines 242 and 244, and there is an obstacle 220 in front of it (to the right). By scanning along the reference line to generate a hard-constraint pipeline ( Figure 2 defined by lines 252 and 254 in ), the hard-constraint pipeline represents the first path region where the host vehicle can safely pass. Based on the hard-constraint pipeline defined by lines 252 and 254, a soft-constraint pipeline is generated by adding lateral and longitudinal buffers ( Figure 2 defined by lines 262 and 264 in ), and the soft-constraint pipeline represents the second path region where the host vehicle can safely pass. As can be seen from Figure 2 , the second path region is smaller than the first path region. Finally, based on the soft-constraint pipeline defined by lines 262 and 264, considering the influence of the steering limit of the host vehicle and the predicted trajectory of the obstacle, a reference pipeline is generated ( Figure 2 defined by lines 272 and 274 in ), and the reference pipeline represents the third path region where the host vehicle can safely pass, and this third path region is smaller than the second path region. In this way, when solving the optimization problem of model predictive control in subsequent steps, the center line of the reference pipeline ( Figure 2 not shown in ) is used as the reference value of this optimization problem, thereby generating the trajectory 230 of the host vehicle.

[0047] In one or more embodiments, the reference line is generated depending on the lane boundaries sensed by the sensor. For example, the reference line is generated based on the sensed lane boundaries using a path planning algorithm (such as polynomial fitting). The reference line is the path that the vehicle needs to follow during driving, usually represented as a smooth curve or polyline. In path planning, the reference line provides the driving direction and path benchmark for the vehicle. It ensures that the vehicle travels in the correct lane and serves as the basis for generating the hard-constraint pipeline, the soft-constraint pipeline, and the reference pipeline.

[0048] During driving, the vehicle updates the sensor data in real time and dynamically adjusts the reference line according to environmental changes. For example, when an obstacle or traffic congestion is detected in the front lane, the system can re-plan the reference line to guide the vehicle to change lanes in advance or take other avoidance measures.

[0049] In addition, since the model predictive control algorithm only considers the optimization problem within a limited time range at each time step, it may cause the vehicle to fall into a local optimal solution rather than a global optimal solution. Therefore, in one embodiment, the pipeline-based model predictive control trajectory planning method 1000 depends on the reference line provided by the global path planning to generate the hard-constraint pipeline, the soft-constraint pipeline, and the reference pipeline, so as to obtain the global optimal solution.

[0050] In step S130, when solving the optimization problem of model predictive control, the center line of the reference pipeline is used as the reference value of the optimization problem, thereby generating the trajectory of the host vehicle.

[0051] In the MPC optimization problem, the center line of the reference pipeline provides the ideal path that the vehicle should follow. By using the center line of the reference pipeline as the reference value of the optimization problem, the optimization algorithm will strive to make the planned trajectory of the vehicle as close as possible to this center line while satisfying various constraints.

[0052] In one embodiment, the goal of the MPC optimization problem is to minimize an objective function (also known as the "cost function"), which consists of the following parts:

[0053] (1) Position error term: Measures the deviation between the current position of the vehicle and the reference value (center line of the reference pipeline). This ensures that the trajectory of the vehicle is as close as possible to the reference path;

[0054] (2) Steering rate change term: Measures the change in the steering rate of the vehicle to ensure the smoothness of the steering action and avoid drastic steering changes;

[0055] (3) Environment constraint term: Considers the distance between the vehicle and the surrounding environment (such as obstacles) to ensure that the vehicle travels within a safe range;

[0056] (4) Acceleration constraint term: Measures the change in the vehicle's acceleration to ensure that the vehicle's acceleration is within a safe and comfortable range.

[0057] In the optimization problem, in addition to the objective function, a series of constraints need to be satisfied to ensure that the movement of the vehicle is safe and feasible. Specifically, these constraints may include:

[0058] (1) Kinematic constraints: Ensure that the movement of the vehicle conforms to its kinematic model, that is, the change in the vehicle's motion state conforms to physical laws;

[0059] (2) Steering actuator limitations: Ensure that the steering angle and steering rate of the vehicle are within the capabilities of the actuator to avoid exceeding the physical limitations of the steering system;

[0060] (3) Acceleration constraints: Ensure that the acceleration and deceleration of the vehicle are within a safe and comfortable range to avoid excessive acceleration or deceleration; and

[0061] (4) Environment constraints: Ensure that the vehicle maintains a sufficient safe distance from surrounding obstacles to avoid collisions.

[0062] By solving the above optimization problem, the optimal control input of the vehicle over a future period can be obtained, thereby generating the vehicle's trajectory. This trajectory will be as close as possible to the centerline of the reference pipeline while satisfying all constraints, ensuring that the vehicle can maintain a safe distance during driving and conform to the vehicle's dynamic characteristics.

[0063] In one embodiment, the objective function can be constructed as follows:

[0064]

[0065] In the above objective function, the position error term includes: w de [k]·(x[k][1]-de ref [k]) 2 +w pe [k]·(x[k][2]-pe ref [k]) 2 , where w de [k] is a weight coefficient, which is used to balance the importance of the longitudinal position error in the objective function. For example, a larger weight value indicates that the longitudinal position error is more important in the optimization problem, and the optimization algorithm will work harder to reduce the longitudinal position error. x[k][1] represents the actual longitudinal position of the vehicle, that is, the longitudinal position of the vehicle at time step k; de ref [k] represents the desired longitudinal position, that is, the longitudinal position that the vehicle should reach at time step k, which can be determined based on the centerline of the reference pipeline. Similarly, w pe [k] is a weight coefficient, which is used to balance the importance of the lateral position error in the objective function. For example, a larger weight value indicates that the lateral position error is more important in the optimization problem, and the optimization algorithm will work harder to reduce the lateral position error. x[k][2] represents the actual lateral position of the vehicle, that is, the lateral position of the vehicle at time step k; pe ref [k] represents the desired lateral position, that is, the lateral position that the vehicle should reach at time step k, which can be determined based on the centerline of the reference pipeline.

[0066] The steering rate change term is: w rotate [k]·(δ[k]-δ[k-1]-rwa_rate_ref[k]) 2 , which is used to measure the deviation between the steering rate change of the vehicle at time step k and the desired steering rate change. Among them, w rotate[k] is the weight coefficient, which is used to balance the importance of the steering rate change term in the objective function. For example, a larger weight value indicates that the steering rate change is more important in the optimization problem, and the optimization algorithm will work harder to reduce the steering rate change. δ[k] represents the actual steering angle of the vehicle at time step k, and δ[k - 1] represents the actual steering angle of the vehicle at time step k - 1. rwa_rate_ref[k] represents the desired steering rate change, which is used as a reference value for the steering rate change of the vehicle at time step k and is used to calculate the deviation between the actual steering rate change and the expected value. In one or more embodiments, rwa_rate_ref[k] can be determined based on the center line of the reference pipeline and / or the vehicle dynamics model.

[0067] Environmental constraint term Used to measure the distance constraint between the vehicle and the surrounding environment (such as obstacles) at time step k. w env 、 And Are all weight coefficients, which are respectively used to balance the importance of the environmental constraint term, the left soft environmental constraint term, and the right soft environmental constraint term in the objective function. S env [k] is a slack variable, which represents the distance relaxation between the vehicle and the surrounding environment (such as obstacles) at time step k and is used to handle soft constraints, allowing slight violation of the constraints in infeasible cases. Similarly, Is a slack variable, which represents the distance relaxation between the vehicle and the left environment at time step k and is used to handle the left soft constraint, allowing slight violation of the constraint in infeasible cases. Is a slack variable, which represents the distance relaxation between the vehicle and the right environment at time step k and is used to handle the right soft constraint, allowing slight violation of the constraint in infeasible cases.

[0068] The acceleration constraint term is Where And Are all weight coefficients, which are respectively used to balance the importance of the minimum acceleration constraint term and the maximum acceleration constraint term in the objective function. S accel_min [k] and S accel_max [k] are both slack variables, which respectively represent the relaxation of the minimum acceleration constraint and the maximum acceleration constraint of the vehicle at time step k and are used to handle soft constraints, allowing slight violation of the constraints in infeasible cases.

[0069] In one embodiment, the constraint conditions include:

[0070] (1) Kinematic constraints, which are used to ensure that the motion state of the vehicle conforms to its kinematic model. The model is, for example: x[k] = A[k - 1] · x[k - 1] + Bδ [k - 1]·(δ[k - 1] + δ[k]) + B k [k - 1]·(κ[k - 1] + κ[k]),

[0071] Where x[k] represents the state vector of the vehicle at time step k, usually including position, speed, steering angle, etc.; A[k - 1] is the state transition matrix, describing the linear change of the vehicle state; B δ [k - 1] and B κ [k - 1] are the input matrices, which describe the influence of the control input on the vehicle state; δ[k - 1] and δ[k] represent the steering angles at time steps k - 1 and k respectively; κ[k - 1] and κ[k] represent the curvatures at time steps k - 1 and k respectively;

[0072] (2) Steering actuator limit, used to ensure that the steering angle of the vehicle is within the capabilities of the actuator, for example: -δ max [k] ≤ δ[k] ≤ δ max [k], where δ max [k] represents the maximum steering angle at time step k;

[0073] (3) Steering rate change constraint, used to limit the rate of change of the vehicle's steering angle to ensure the smoothness of the steering action, for example: slew_min[k] ≤ δ[k] - δ[k - 1] ≤ slew_max[k], where slew_min[k] and slew_max[k] represent the minimum and maximum steering rate changes at time step k respectively;

[0074] (4) Acceleration constraint, used to ensure that the acceleration of the vehicle is within a safe and feasible range, for example: accel min [1] - S accel_min [1] ≤ δ[1] - δ[0] - slew0 ≤ accel max [1] + S accel_max [1], and accel min [k + 1] - S accel_min [k + 1] ≤ δ[k + 1] - 2·δ[k] + δ[k - 1] ≤ accel max [k + 1] + S accel_max [k + 1], where, accel min [k] and accel max [k] represent the minimum and maximum accelerations at time step k respectively, S accel_min [k] and S accel_max [k] represent the acceleration constraint relaxation variables at time step k respectively, and slew0 represents the reference value of the steering rate change or acceleration at the initial time (time step 0);

[0075] (5) Environmental constraints, which are used to ensure that the vehicle maintains a safe distance from the surrounding environment (such as obstacles), for example: H env ·x[k] ≤ G env (k) + S env (k), where H env is the environmental constraint matrix, G env (k) is the environmental constraint vector at time step k, and S env (k) is the environmental constraint slack variable at time step k;

[0076] (6) Left - hand soft environmental constraints, which are used to ensure that the vehicle maintains a safe distance from the left - hand environment and allows slight violation of the constraints, for example: H env_soft_left ·x[k] ≤ G env_soft_left (k) + S env_soft_left (k), where H env_soft_left is the left - hand soft environmental constraint matrix, G env_soft_left (k) is the left - hand soft environmental constraint vector at time step k, and S env_soft_left (k) is the left - hand soft environmental constraint slack variable at time step k;

[0077] (7) Right - hand soft environmental constraints, which are used to ensure that the vehicle maintains a safe distance from the right - hand environment and allows slight violation of the constraints, for example: H env_soft_right ·x[k] ≤ G env_soft_right (k) + S env_soft_right (k), where H env_soft_right is the right - hand soft environmental constraint matrix, G env_soft_right (k) is the right - hand soft environmental constraint vector at time step k, and S env_soft_right (k) is the right - hand soft environmental constraint slack variable at time step k.

[0078] In the above - mentioned embodiments, the objective function of the optimization problem incorporates position error, steering rate change, environmental constraints, acceleration constraints, etc. By balancing different objectives through reasonable weight coefficients, multi - objective optimization is achieved. For example, by optimizing the steering rate change and acceleration change, the frequent steering of the vehicle during driving is reduced, enhancing the comfort experience of passengers. Additionally, when solving the optimization problem, slack variables are introduced to allow slight violation of the constraints in infeasible cases, ensuring that a feasible solution can be found for the optimization problem.

[0079] In addition, those skilled in the art can easily understand that the pipeline-based model predictive control trajectory planning method 1000 provided by one or more of the above embodiments of the present application can be implemented by a computer program. For example, the computer program is included in a computer program product, and when the computer program is executed by a processor, it implements the pipeline-based model predictive control trajectory planning method 1000 of one or more embodiments of the present application. For another example, when a computer-readable storage medium (such as a USB flash drive) storing the computer program is connected to a computer, running the computer program can execute the pipeline-based model predictive control trajectory planning method 1000 of one or more embodiments of the present application.

[0080] The above trajectory planning method 1000 can be applied to various autonomous driving functions. In one embodiment, the above trajectory planning method 1000 can be applied to an adaptive cruise control function. For example, the method of adaptive cruise control may include: receiving the environmental information of the host vehicle from multiple sensors; based on the environmental information, executing the pipeline-based model predictive control trajectory planning method 1000; and providing the planned trajectory to the lateral and longitudinal actuators for lateral and longitudinal control of the host vehicle.

[0081] Refer to Figure 3 , which shows a schematic structural diagram of an electronic device 3000 according to an embodiment of the present application. As Figure 3 shown, the electronic device 3000 includes a memory 310 and a processor 320, and a computer program is stored on the memory 310. In one embodiment, the processor 320 executes the computer program to implement the following functions or steps: receiving the environmental information of the host vehicle from multiple sensors; based on the environmental information, executing the pipeline-based model predictive control trajectory planning method 1000 as described above; and providing the planned trajectory to the lateral and longitudinal actuators for controlling the host vehicle. In one or more embodiments, the electronic device may be a domain controller, a camera, or a radar.

[0082] In summary, in the pipeline-based model predictive control trajectory planning solution of the embodiments of the present application, the influence of vehicle dynamics constraints and obstacle prediction trajectories is considered during the process of constructing a pipeline (i.e., a tube, such as a reference pipeline), so that while narrowing the feasible region of the optimization problem, it is ensured that the vehicle will not collide with obstacles during driving, thereby enabling the rapid generation of a collision-free trajectory that conforms to the dynamics constraints.

[0083] In one or more embodiments, by generating a hard constraint pipeline, a soft constraint pipeline, and a reference pipeline, the feasible region of the optimization problem is gradually narrowed, and the search space of the optimization problem is reduced, thereby improving the solution efficiency. A smaller feasible region means a lower complexity of the optimization problem, and the solver can more efficiently find the optimal path that satisfies all constraint conditions, meeting the real-time requirements.

[0084] The above examples mainly illustrate the trajectory planning scheme of the embodiments of the present application. Although only some of the embodiments of the present application have been described, those of ordinary skill in the art should understand that the present application can be implemented in many other forms without departing from its gist and scope. Therefore, the examples and embodiments shown are regarded as illustrative rather than restrictive, and the present application may cover various modifications and substitutions without departing from the spirit and scope of the present application as defined by the various claims.

Claims

1. A pipeline-based model predictive control trajectory planning method, characterized in that: The method comprises: Collecting environmental information around the vehicle, wherein the environmental information includes information related to obstacles; generating a reference conduit representing a drivable space of the ego vehicle based at least on the steering constraints of the ego vehicle and the predicted trajectory of the obstacle; and When solving the optimization problem of the model predictive control, the center line of the reference pipeline is used as a reference value of the optimization problem, thereby generating a trajectory of the ego vehicle so that the ego vehicle can travel along the generated trajectory.

2. The method of claim 1, wherein generating a reference pipeline representing the drivable space of the ego vehicle in combination with the steering restriction of the ego vehicle and the predicted trajectory of the obstacle comprises: Scanning along the reference line to generate a hard-constrained pipeline, where the hard-constrained pipeline represents a first path area where the vehicle can safely pass, where the first path area is defined by a road boundary in a physical sense; On the basis of the hard constraint pipeline, a soft constraint pipeline is generated by adding lateral and longitudinal buffer zones, wherein the soft constraint pipeline represents a second path area where the vehicle can safely pass, and the second path area is smaller than the first path area; as well as A reference pipeline is generated based on the soft constraint pipeline by considering the influence of the steering restriction of the ego vehicle and the predicted trajectory of the obstacle. The reference pipeline represents a third path area in which the ego vehicle can pass safely, and the third path area is smaller than the second path area.

3. The method of claim 2, wherein: The environmental information also includes a lane boundary sensed by a sensor, and the reference line is generated depending on the lane boundary.

4. The method according to claim 1 or 2, wherein: The obstacle-related information includes the position and speed of the obstacle, and the predicted trajectory of the obstacle is determined based on the position and speed.

5. The method of claim 1, wherein: The optimization problem of the model predictive control is to minimize the objective function while satisfying the constraints.

6. The method of claim 5, wherein: The objective function is composed of a position error term, a steering rate change term, an environmental constraint term, and an acceleration constraint term, and the constraint conditions include dynamic constraints, steering actuator limitations, acceleration constraints, and environmental constraints.

7. The method of claim 1, wherein: The environmental information is received from a plurality of sensors, and the method further comprises: The generated trajectory of the ego vehicle is provided to the lateral and longitudinal actuators for controlling the ego vehicle.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

9. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the method according to any one of claims 1 to 7.

10. The electronic device according to claim 9, wherein: The electronic device is a domain controller, a camera or a radar.

Images