Vehicle queue cooperative avoidance trajectory planning method under freeway accident scene
By using cloud-based reinforcement learning methods, collaborative planning trajectories are generated based on traffic flow dynamics and vehicle platoon status, solving the safety response problem of vehicle platoons under sudden accidents and realizing collaborative avoidance of vehicle platoons in highway accident scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA AGRI UNIV
- Filing Date
- 2026-03-13
- Publication Date
- 2026-07-14
AI Technical Summary
In mixed traffic flow on highways, vehicle queues struggle to respond quickly and safely in the event of a sudden accident. Traditional methods and conventional reinforcement learning algorithms have weak generalization capabilities, leading to the failure of cooperative collision avoidance.
A two-layer optimized meta-reinforcement learning method based on powerful cloud computing and beyond-line-of-sight perception is adopted. By acquiring traffic flow dynamics and fleet status, a state vector is constructed. Iterative training is performed using a pre-constructed vehicle kinematics model and a meta-reinforcement learning objective function to generate a collaboratively planned trajectory, enabling rapid online adaptation.
In the event of an emergency, it enables coordinated safety avoidance of vehicle platoons, improves the rapid adaptability and safety of decision-making, and ensures the safe passage of vehicle platoons.
Smart Images
Figure CN122392349A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of intelligent transportation and autonomous driving technology, and in particular to a method for planning vehicle platoon collaborative avoidance trajectories in highway accident scenarios. Background Technology
[0002] In a fully connected vehicle (CAV) environment, multi-vehicle platooning can effectively improve road traffic efficiency and reduce energy consumption. However, in mixed traffic flows on highways (including human-driven HDVs), when a sudden accident occurs ahead of the platoon, traditional vehicle control methods or conventional reinforcement learning algorithms often struggle to respond quickly and safely in unknown, dynamic accident scenarios. Conventional algorithms also exhibit weak generalization ability when faced with unseen obstacle distributions or traffic flow densities, easily leading to cooperative collision avoidance failures.
[0003] Therefore, there is an urgent need for a collaborative decision-making and planning method for vehicle platooning that can leverage the powerful computing capabilities and beyond-line-of-sight perception capabilities of the cloud, and has the ability to quickly adapt to new scenarios. Summary of the Invention
[0004] This invention provides a method for planning the collaborative avoidance trajectory of vehicle platoons in highway accident scenarios, in order to solve the problem of collaborative safety avoidance of vehicle platoons under sudden accidents.
[0005] A first aspect of the present invention provides a method for planning a vehicle platoon cooperative avoidance trajectory in a highway accident scenario, comprising the following steps: Obtain traffic flow dynamics and convoy driving status in the accident area to construct a state vector; A pre-built vehicle kinematics model is used to describe the state transition process of any vehicle in the fleet at any time, serving as a physical constraint for training and planning. The pre-constructed meta-reinforcement learning objective function based on two-layer optimization is iteratively trained using the state vector and the physical constraints until the preset multi-view composite reward function converges, thus obtaining the trained meta-policy model. When a sudden accident is detected, the state vector is used to quickly adapt online to the trained meta-policy model to generate a collaboratively planned trajectory.
[0006] Optionally, the specific expression of the physical constraint is:
[0007] in, For the coordinates of the vehicle's center of gravity, For the first The car is The longitudinal velocity at time t, For decision-making time steps, For the first The car is The heading angle at that moment, For the front wheel steering angle, This refers to the vehicle's wheelbase. For the first The car is The longitudinal acceleration at time t.
[0008] Optionally, the meta-reinforcement learning objective function based on two-layer optimization includes an offline meta-training phase and an online rapid adaptation phase, wherein, The offline meta-training phase is used to construct a task distribution that includes multiple accident types; The online rapid adaptation phase is used to update the gradient or inference steps based on the real-time observation data uploaded by the roadside equipment and the prior knowledge obtained by meta-learning, so as to adjust the task distribution containing multiple accident types into a collaborative planning trajectory for the current specific accident scenario.
[0009] Optionally, the specific expression of the meta-reinforcement learning objective function based on two-layer optimization is as follows:
[0010]
[0011] In the formula, These are the initial parameters for the meta-policy. The objective function value of the meta-policy. For task distribution Upsampling specific incident tasks The mathematical expectation, For adapted task-specific strategies downsampling trajectory The mathematical expectation, H represents the planning time domain. As a discount factor, For the current decision-making step, For a multi-objective composite reward function, Let be the state vector at time t. Let be the action vector at time t. Parameters specific to the adapted task. The learning rate for the inner layer update. For the initial parameters of the meta-policy gradient operator, For meta-strategy downsampling trajectory The mathematical expectation.
[0012] Optionally, the specific expression of the preset multi-view composite reward function is as follows:
[0013] in, For a multi-objective composite reward function, The weighting coefficient for the safety item. For security purposes, The weighting coefficients for the efficiency term. For efficiency, This is the weighting coefficient for the comfort item. For comfort items, The weighting coefficients for the synergy term. It is a synergistic item.
[0014] A second aspect of the present invention provides a planning device for a vehicle platoon cooperative avoidance trajectory in a highway accident scenario, comprising: The state vector construction module is used to obtain the traffic flow dynamics and convoy driving status in the accident area in order to construct state vectors; The state transition description module is used to describe the state transition process of any vehicle in the fleet at any time using a pre-built vehicle kinematic model, as a physical constraint for training and planning. The training module is used to iteratively train a pre-constructed meta-reinforcement learning objective function based on bi-layer optimization using the state vector and the physical constraints until the preset multi-view composite reward function converges, thus obtaining the trained meta-policy model. The generation module is used to generate a collaborative planning trajectory by rapidly adapting the state vector to the trained meta-policy model when a sudden accident is detected.
[0015] A third aspect of the present invention provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the vehicle platoon cooperative avoidance trajectory planning method in a highway accident scenario as described in the above embodiments.
[0016] A fourth aspect of the present invention provides a computer program product, which, when executed by a processor, implements the above-described method for planning vehicle platoon cooperative avoidance trajectories in a highway accident scenario.
[0017] A fifth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for planning vehicle platoon cooperative avoidance trajectories in a highway accident scenario.
[0018] The proposed method for planning collaborative avoidance trajectories for vehicle platoons in highway accident scenarios, as described in this invention, firstly, integrates roadside perception with cloud computing to obtain real-time traffic flow dynamics and platoon driving status in the accident area, and features the accident scenario into a task distribution based on meta-reinforcement learning; secondly, it constructs an integrated decision-making and planning model based on meta-reinforcement learning, utilizes computing power advantages in the cloud for multi-task meta-training, and rapidly adapts online according to the characteristics of the current sudden accident to generate a collaborative avoidance trajectory; finally, it distributes the planned safe trajectory to the vehicle, and the vehicle controller tracks the trajectory in conjunction with vehicle dynamics constraints to achieve collaborative safety avoidance in sudden accidents.
[0019] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0020] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a schematic diagram illustrating the specific execution of a vehicle platoon collaborative avoidance trajectory planning method in a highway accident scenario according to an embodiment of the present invention; Figure 2 A flowchart illustrating a method for planning a vehicle platoon cooperative avoidance trajectory in a highway accident scenario, according to an embodiment of the present invention; Figure 3 A block diagram of a vehicle platoon cooperative avoidance trajectory planning device in a highway accident scenario according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention.
[0021] Explanation of reference numerals in the attached figures: 30-Planning device for vehicle platoon collaborative avoidance trajectory in highway accident scenarios, 301-State vector construction module, 302-State transition description module, 303-Training planning module, 401-Memory, 402-Processor, 403-Communication interface. Detailed Implementation
[0022] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.
[0023] The following describes a method for planning vehicle platoon collaborative avoidance trajectories in a highway accident scenario according to an embodiment of the present invention, with reference to the accompanying drawings.
[0024] It should be noted that, as Figure 1 As shown, the research scenario of this invention is based on a vehicle-road-cloud collaborative architecture. For example, it is set as follows: on a two-lane highway, three autonomous vehicles form a convoy, and there is traffic flow formed by other human-driven vehicles (HDVs) on the highway. The road scenario is equipped with infrastructure under the vehicle-road-cloud architecture, including a cloud server, roadside units (RSUs), and on-board units (OBUs). When a sudden accident occurs ahead of the lane where the vehicle convoy is located, the convoy coordinates with each other through the on-board units (OBUs), and the convoy communicates with the cloud server through the RSUs.
[0025] The cloud computing center, acting as the "brain," acquires traffic flow and vehicle information and deploys meta-reinforcement learning algorithms to process high-dimensional perception data, train and infer models, and generate collaborative avoidance strategies that include speed and path information for sudden accidents. The roadside unit (RSU) provides beyond-line-of-sight perception capabilities, monitors sudden accidents beyond the line of sight in front of the vehicle queue, and uploads environmental status and other information to the cloud. The queues coordinate with each other through the onboard unit (OBU), and the vehicle-side execution layer receives the planned trajectory from the cloud through the OBU and controls the vehicle chassis actuators (accelerator, brake, steering) to complete specific avoidance actions.
[0026] Figure 2 This is a flowchart illustrating a method for planning a vehicle platoon's collaborative avoidance trajectory in a highway accident scenario, as provided in an embodiment of the present invention.
[0027] like Figure 2 As shown, the planning method for the vehicle platoon cooperative avoidance trajectory in this highway accident scenario is implemented in a cloud computing center, and includes the following steps: In step S201, the traffic flow dynamics and vehicle driving status of the accident area are obtained to construct a state vector.
[0028] In some embodiments, the state vector includes vehicle state, road geometry constraints, accident obstacle information, and a latent vector containing characteristics of the accident scenario.
[0029] In actual implementation, the traffic flow dynamics and vehicle driving status of the accident area are integrated to construct a high-dimensional state vector S that includes vehicle status (position, speed, acceleration), road geometric constraints and accident obstacle information, and a potential vector describing the characteristics of the accident scene.
[0030] Furthermore, in order to achieve decoupling and minimization of decision-making and planning, this embodiment of the invention adopts a continuous action space to directly output the avoidance strategy. Therefore, the action vector A is defined as... ,in, For longitudinal acceleration, The front wheel turns, thus completing the decision of "whether to change lanes" and the planning of "how to change lanes" in one step.
[0031] In step S202, a pre-built vehicle kinematics model is used to describe the state transition process of any vehicle in the fleet at any time, as a physical constraint for training and planning.
[0032] In actual implementation, a vehicle kinematics model is used to describe the first vehicle in the platoon. The car is The state transition process at any given moment serves as a physical constraint for cloud-based training and planning, ensuring that the generated planned trajectory conforms to the nonholonomic constraint characteristics of the vehicle. The specific expression is as follows:
[0033] in, For the coordinates of the vehicle's center of gravity, For the first The car is The longitudinal velocity at time t, For decision-making time steps, For the first The car is The heading angle at that moment, For the front wheel steering angle, This refers to the vehicle's wheelbase. For the first The car is The longitudinal acceleration at time t.
[0034] In step S203, the pre-constructed meta-reinforcement learning objective function based on bi-layer optimization is iteratively trained using state vectors and physical constraints until the preset multi-view composite reward function converges, thus obtaining the collaborative planning trajectory.
[0035] In step S204, when a sudden accident is detected, the system performs online rapid adaptation based on the state vector on the trained meta-policy model to generate a collaborative planning trajectory.
[0036] In some embodiments, the meta-reinforcement learning objective function based on bi-layer optimization includes an offline meta-training phase and an online rapid adaptation phase, wherein, The offline meta-training phase is used to build a task distribution that includes multiple accident types; The online rapid adaptation phase is used to update gradients or inference steps based on real-time observation data uploaded by roadside equipment and prior knowledge obtained through meta-learning, so as to adjust the task distribution containing multiple accident types into a collaborative planning trajectory for the current specific accident scenario.
[0037] In actual implementation, a meta-reinforcement learning objective function based on two-layer optimization is first established, and the solution of the cloud-based collaborative avoidance strategy is defined as maximizing the task distribution. The expected cumulative reward includes two processes: inner adaptation (i.e., the online rapid adaptation phase) and outer update (i.e., the offline meta-training phase).
[0038] In the formula, These are the initial parameters for the meta-policy. The objective function value of the meta-policy. For task distribution Upsampling specific incident tasks The mathematical expectation, For adapted task-specific strategies downsampling trajectory The mathematical expectation, where H is the time domain of the planning. As a discount factor, For the current decision-making step, For a multi-objective composite reward function, Let be the state vector at time t. Let be the action vector at time t. Parameters specific to the adapted task.
[0039] In the offline meta-training phase, a task distribution with varying obstacle locations, traffic flow densities, and driver behaviors is constructed in a cloud-based simulation environment. The adapted parameters are then evaluated based on the query set data. The performance of backpropagation updates meta-parameters. This allows the general strategy to maximize the expected cumulative reward across all training tasks and to adapt quickly.
[0040] When a real-world emergency occurs, the online rapid adaptation phase utilizes the limited real-time observation data context uploaded by roadside equipment, along with specific emergency tasks. The task gradient is calculated using the support set data. For the meta-parameters... Perform one or more gradient updates to obtain the adapted task-specific parameters. :
[0041] In the formula, Parameters specific to the adapted task. The learning rate for the inner layer update. For the initial parameters of the meta-policy gradient operator, For meta-strategy downsampling trajectory The mathematical expectation.
[0042] This process requires only a minimal amount of computation to adjust to the optimal avoidance strategy for the current specific accident scenario, ensuring real-time decision-making and safety.
[0043] Furthermore, the pre-constructed meta-reinforcement learning objective function based on bi-layer optimization is iteratively trained using state vectors and physical constraints until the preset multi-view composite reward function converges, thus obtaining the collaborative planning trajectory.
[0044] The specific expression for the preset multi-view composite reward function is as follows:
[0045] In the formula, For a multi-objective composite reward function, The weighting coefficient for the safety item. For security purposes, The weighting coefficients for the efficiency term. For efficiency, This is the weighting coefficient for the comfort item. For comfort items, The weighting coefficients for the synergy term. It is a synergistic item.
[0046] It should be noted that the security item Based on the principle of the potential field method, when the distance between the vehicle and the obstacle... Less than the safety threshold Apply nonlinear penalty at the time:
[0047] Efficiency Item To minimize the vehicle speed relative to the desired cruising speed Deviation:
[0048] Comfort items For: Penalty acceleration With front wheel steering angle change rate High-frequency jitter:
[0049] Synergistic items For: Penalize adjacent vehicles in the convoy i, j Speed difference and distance error between them: .
[0050] In summary, the method for planning collaborative avoidance trajectories of vehicle platoons in highway accident scenarios proposed in this embodiment of the invention firstly acquires real-time traffic flow dynamics and platoon driving status in the accident area based on the fusion of roadside perception and cloud computing, and features the accident scenario into a task distribution of meta-reinforcement learning; secondly, it constructs an integrated decision-making and planning model based on meta-reinforcement learning, performs multi-task meta-training in the cloud using computing power advantages, and rapidly adapts online according to the characteristics of the current sudden accident to generate a collaborative avoidance trajectory; finally, it sends the planned safe trajectory to the vehicle end, and the vehicle end controller tracks the trajectory in combination with vehicle dynamics constraints to achieve collaborative safety avoidance in sudden accidents.
[0051] Next, referring to the accompanying drawings, a planning device for vehicle platoon cooperative avoidance trajectory in a highway accident scenario according to an embodiment of the present invention is described.
[0052] Figure 3 This is a block diagram of a vehicle platoon collaborative avoidance trajectory planning device in a highway accident scenario provided by an embodiment of the present invention.
[0053] like Figure 3 As shown, the planning device 30 for the vehicle platoon cooperative avoidance trajectory in the highway accident scenario includes: a state vector construction module 301, a state transition description module 302, and a training planning module 303.
[0054] The system comprises the following modules: a state vector construction module 301, which acquires traffic flow dynamics and convoy driving status in the accident area to construct state vectors; a state transition description module 302, which uses a pre-built vehicle kinematics model to describe the state transition process of any vehicle in the convoy at any given time, serving as a physical constraint for training and planning; a training module 303, which iteratively trains a pre-built meta-reinforcement learning objective function based on bi-layer optimization using the state vectors and the physical constraints until the preset multi-view composite reward function converges, resulting in a trained meta-policy model; and a generation module 304, which, upon detecting a sudden accident, performs online rapid adaptation based on the state vectors on the trained meta-policy model to generate a collaboratively planned trajectory.
[0055] In some embodiments, the state vector includes vehicle state, road geometry constraints, accident obstacle information, and a latent vector containing characteristics of the accident scenario.
[0056] In some embodiments, the specific expression of the physical constraint is:
[0057] in, For the coordinates of the vehicle's center of gravity, For the first The car is The longitudinal velocity at time t, For decision-making time steps, For the first The car is The heading angle at that moment, For the front wheel steering angle, This refers to the vehicle's wheelbase. For the first The car is The longitudinal acceleration at time t.
[0058] In some embodiments, the meta-reinforcement learning objective function based on bi-layer optimization includes an offline meta-training phase and an online rapid adaptation phase, wherein, The offline meta-training phase is used to build a task distribution that includes multiple accident types; The online rapid adaptation phase is used to update gradients or inference steps based on real-time observation data uploaded by roadside equipment and prior knowledge obtained through meta-learning, so as to adjust the task distribution containing multiple accident types into a collaborative planning trajectory for the current specific accident scenario.
[0059] In some embodiments, the specific expression of the meta-reinforcement learning objective function based on bilayer optimization is:
[0060]
[0061] In the formula, These are the initial parameters for the meta-policy. The objective function value of the meta-policy. For task distribution Upsampling specific incident tasks The mathematical expectation, For adapted task-specific strategies downsampling trajectory The mathematical expectation, where H is the time domain of the planning. As a discount factor, For the current decision-making step, For a multi-objective composite reward function, Let be the state vector at time t. Let be the action vector at time t. Parameters specific to the adapted task. The learning rate for the inner layer update. For the initial parameters of the meta-policy gradient operator, For meta-strategy downsampling trajectory The mathematical expectation.
[0062] In some embodiments, the specific expression of the preset multi-view composite reward function is as follows:
[0063] in, For a multi-objective composite reward function, The weighting coefficient for the safety item. For security purposes, The weighting coefficients for the efficiency term. For efficiency, This is the weighting coefficient for the comfort item. For comfort items, The weighting coefficients for the synergy term. It is a synergistic item.
[0064] It should be noted that the explanation of the above-mentioned embodiment of the vehicle platoon cooperative avoidance trajectory planning method in highway accident scenarios also applies to the vehicle platoon cooperative avoidance trajectory planning device in highway accident scenarios in this embodiment, and will not be repeated here.
[0065] The vehicle platoon collaborative avoidance trajectory planning device for highway accident scenarios proposed in this embodiment of the invention firstly acquires the traffic flow dynamics and platoon driving status of the accident area in real time based on the fusion of roadside perception and cloud, and features the accident scenario into a task distribution of meta-reinforcement learning; secondly, it constructs an integrated decision planning model based on meta-reinforcement learning, performs multi-task meta-training in the cloud using computing power advantages, and rapidly adapts online according to the characteristics of the current sudden accident to generate a collaborative avoidance trajectory; finally, it sends the planned safe trajectory to the vehicle end, and the vehicle end controller tracks the trajectory in combination with vehicle dynamics constraints to achieve collaborative safety avoidance under sudden accidents.
[0066] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.
[0067] The electronic device may include: a memory 401, a processor 402, and a computer program stored on the memory 401 and capable of running on the processor 402.
[0068] When the processor 402 executes the program, it implements the vehicle platoon collaborative avoidance trajectory planning method provided in the above embodiment for highway accident scenarios.
[0069] Furthermore, electronic devices also include: Communication interface 403 is used for communication between memory 401 and processor 402.
[0070] The memory 401 is used to store computer programs that can run on the processor 402.
[0071] Memory 401 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.
[0072] If the memory 401, processor 402, and communication interface 403 are implemented independently, then the communication interface 403, memory 401, and processor 402 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0073] Optionally, in a specific implementation, if the memory 401, processor 402, and communication interface 403 are integrated on a single chip, then the memory 401, processor 402, and communication interface 403 can communicate with each other through an internal interface.
[0074] Processor 402 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.
[0075] This invention also provides a computer program product, which, when executed by a processor, implements the above-mentioned method for planning vehicle platoon collaborative avoidance trajectories in a highway accident scenario.
[0076] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for planning vehicle platoon cooperative avoidance trajectories in a highway accident scenario.
[0077] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0078] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.
[0079] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.
[0080] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.
[0081] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.
[0082] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.
[0083] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.
[0084] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.
Claims
1. A method for planning vehicle platoon cooperative avoidance trajectories in highway accident scenarios, characterized in that, When applied to the cloud, the following steps are included: Obtain traffic flow dynamics and convoy driving status in the accident area to construct a state vector; A pre-built vehicle kinematics model is used to describe the state transition process of any vehicle in the fleet at any time, serving as a physical constraint for training and planning. The pre-constructed meta-reinforcement learning objective function based on two-layer optimization is iteratively trained using the state vector and the physical constraints until the preset multi-view composite reward function converges, thus obtaining the trained meta-policy model. When a sudden accident is detected, the state vector is used to quickly adapt online to the trained meta-policy model to generate a collaboratively planned trajectory.
2. The method for planning vehicle platoon cooperative avoidance trajectories in highway accident scenarios according to claim 1, characterized in that, The state vector includes the fleet state, road geometry constraints, accident obstacle information, and a potential vector containing features describing the accident scene.
3. The method for planning vehicle platoon cooperative avoidance trajectories in highway accident scenarios according to claim 1, characterized in that, The specific expression for the physical constraint is: in, For the coordinates of the vehicle's center of gravity, For the first The car is The longitudinal velocity at time t, For decision-making time steps, For the first The car is The heading angle at that moment, For the front wheel steering angle, This refers to the vehicle's wheelbase. For the first The car is The longitudinal acceleration at time t.
4. The method for planning vehicle platoon cooperative avoidance trajectories in highway accident scenarios according to claim 1, characterized in that, The meta-reinforcement learning objective function based on bi-layer optimization includes an offline meta-training phase and an online rapid adaptation phase, wherein... The offline meta-training phase is used to construct a task distribution that includes multiple accident types; The online rapid adaptation phase is used to update the gradient or inference steps based on the real-time observation data uploaded by the roadside equipment and the prior knowledge obtained by meta-learning, so as to adjust the task distribution containing multiple accident types into a collaborative planning trajectory for the current specific accident scenario.
5. The method for planning vehicle platoon cooperative avoidance trajectories in highway accident scenarios according to claim 1 or 4, characterized in that, The specific expression for the meta-reinforcement learning objective function based on bi-layer optimization is as follows: In the formula, These are the initial parameters for the meta-policy. The objective function value of the meta-policy. For task distribution Upsampling specific incident tasks The mathematical expectation, For task-specific strategies after adaptation downsampling trajectory The mathematical expectation, where H is the time domain of the planning. As a discount factor, For the current decision-making step, For a multi-objective composite reward function, Let be the state vector at time t. Let be the action vector at time t. Parameters specific to the adapted task. The learning rate for the inner layer update. For the initial parameters of the meta-policy gradient operator, For meta-strategy downsampling trajectory The mathematical expectation.
6. The method for planning vehicle platoon cooperative avoidance trajectories in highway accident scenarios according to claim 1, characterized in that, The specific expression of the preset multi-view composite reward function is as follows: in, For a multi-objective composite reward function, The weighting coefficient for the safety item. For security purposes, The weighting coefficients for the efficiency term. For efficiency, This is the weighting coefficient for the comfort item. For comfort items, The weighting coefficients for the synergy term. It is a synergistic item.
7. A planning device for vehicle platoon cooperative avoidance trajectory in a highway accident scenario, characterized in that, include: The state vector construction module is used to obtain the traffic flow dynamics and convoy driving status in the accident area in order to construct state vectors; The state transition description module is used to describe the state transition process of any vehicle in the fleet at any time using a pre-built vehicle kinematic model, as a physical constraint for training and planning. The training module is used to iteratively train a pre-constructed meta-reinforcement learning objective function based on bi-layer optimization using the state vector and the physical constraints until the preset multi-view composite reward function converges, thus obtaining the trained meta-policy model. The generation module is used to generate a collaborative planning trajectory by rapidly adapting the state vector to the trained meta-policy model when a sudden accident is detected.
8. An electronic device, characterized in that, include: The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a method for planning a vehicle platoon cooperative avoidance trajectory in a highway accident scenario as described in any one of claims 1-6.
9. A computer program product, characterized in that, When executed by a processor, the computer program / instruction implements the method for planning vehicle platoon cooperative avoidance trajectories in highway accident scenarios as described in any one of claims 1-6.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the method for planning vehicle platoon cooperative avoidance trajectories in highway accident scenarios as described in any one of claims 1-6.