Trajectory generation method and device, vehicle, and storage medium

Abstract

The application provides a trajectory generation method and device, a vehicle and a storage medium, and is applied to an edge server. The method comprises the following steps: in the case that a first vehicle enters a road intersection area, acquiring first vehicle state data of the first vehicle and second vehicle state data of at least one second vehicle located in the road intersection area; constructing a kinematic model corresponding to the first vehicle based on the first vehicle state data, wherein the kinematic model is used to represent a state transition relationship of the first vehicle at multiple future time points; constructing a target cost function corresponding to the first vehicle based on the kinematic model; generating a target trajectory sequence of the first vehicle at the multiple future time points based on the target cost function and the second vehicle state data; and sending the target trajectory sequence to the first vehicle. The application solves the technical problem of low safety of a vehicle when driving in a road intersection area in the related art.

Description

Technical Field

[0001] This application relates to the field of vehicles, and more specifically, to a trajectory generation method, apparatus, vehicle, and storage medium. Background Technology

[0002] In related technologies, the control of vehicles during ramp merging often relies on single-vehicle intelligent sensing or centralized cloud computing. However, due to the limited detection range of onboard sensors and the high latency of remote cloud computing, the accuracy of vehicle control is low, resulting in lower safety for vehicles traveling in road intersections.

[0003] There is currently no good solution to the above problems. Summary of the Invention

[0004] This application provides a trajectory generation method, apparatus, vehicle, and storage medium to at least address the technical problem of low vehicle safety when driving in road intersections in related technologies.

[0005] According to one aspect of the embodiments of this application, a trajectory generation method is provided, applied to an edge server, comprising: upon detecting that a first vehicle has entered a road intersection area, acquiring first vehicle state data of the first vehicle and second vehicle state data of at least one second vehicle located in the road intersection area, wherein the road intersection area represents the area where a main road and a ramp intersect; constructing a kinematic model corresponding to the first vehicle based on the first vehicle state data, wherein the kinematic model represents the state transition relationship of the first vehicle at multiple future time points; constructing a target cost function corresponding to the first vehicle based on the kinematic model, wherein the target cost function quantifies the comprehensive cost of the first vehicle in a preset time domain; generating a target trajectory sequence of the first vehicle at multiple future time points based on the target cost function and the second vehicle state data; and sending the target trajectory sequence to the first vehicle.

[0006] Furthermore, based on the state data of the first vehicle, a kinematic model corresponding to the first vehicle is constructed, including: determining the state vector and control vector corresponding to the first vehicle based on the state data of the first vehicle, wherein the state vector is used to represent the kinematic state of the first vehicle at the current moment, and the control vector is used to represent the control input of the first vehicle at the current moment; determining the sampling time interval; and constructing a kinematic model based on the state vector, the control vector, and the sampling time interval.

[0007] Furthermore, the first vehicle state data includes at least: vehicle position data, vehicle speed, heading angle, acceleration, and steering wheel angle. Based on the first vehicle state data, the corresponding state vector and control vector of the first vehicle are determined, including: determining the state vector based on the vehicle position data, vehicle speed, and heading angle; and determining the control vector based on the acceleration and steering wheel angle.

[0008] Furthermore, based on the objective cost function and the second vehicle state data, a target trajectory sequence of the first vehicle at multiple future times is generated, including: determining the minimization of the objective cost function as the solution objective; solving the objective cost function based on the solution objective and the second vehicle state data to obtain the target trajectory sequence.

[0009] Furthermore, based on the kinematic model, a target cost function corresponding to the first vehicle is constructed, including: based on the kinematic model, the vehicle speed at multiple future times, and the vehicle control vector at multiple future times, determining the cost value corresponding to at least one target cost term, wherein the at least one target cost term is used to represent a quantitative penalty index at at least one dimension during trajectory generation; based on the weight coefficients corresponding to the at least one target cost term, the cost value corresponding to the at least one target cost term is weighted and processed to construct the target cost function.

[0010] Furthermore, based on the kinematic model, the vehicle speed at multiple future moments, and the vehicle's control vector at multiple future moments, the cost value corresponding to at least one target cost term is determined, including: based on the kinematic model, determining the actual distance between the first vehicle and the second vehicle at multiple future moments, and constructing a safety cost term based on the distance difference between the actual distance between the vehicles and the preset safety distance; constructing an efficiency cost term based on the speed difference between the vehicle speed at multiple future moments and the target vehicle speed; and constructing a comfort cost term based on the rate of change of the control vector at multiple future moments.

[0011] Furthermore, the target cost function is solved based on the objective and the second vehicle state data to obtain the target trajectory sequence, including: determining at least one constraint based on the second vehicle state data; and, under the premise of satisfying at least one constraint, recursively solving the target cost function based on the objective to obtain the target trajectory sequence.

[0012] Furthermore, based on the second vehicle state data, at least one constraint is determined, including: determining a safety constraint based on the actual distance between the first vehicle and the second vehicle at multiple future times; determining a kinematic constraint based on the performance limitation parameters of the first vehicle; determining a spatiotemporal constraint based on the geometric topology information of the road intersection area and the merging task requirements; and determining an obstacle constraint based on the second vehicle state data.

[0013] Furthermore, based on the second vehicle state data, obstacle constraints are determined, including: predicting the state of the second vehicle based on the second vehicle state data to obtain the predicted driving area of ​​the second vehicle at multiple future times; and determining obstacle constraints based on the predicted driving area.

[0014] Furthermore, the method also includes: determining the merging point between the main road and the ramp based on road map data; and constructing a road intersection area based on the merging point.

[0015] Furthermore, after sending the target trajectory sequence to the first vehicle, the method further includes: performing update monitoring on the road intersection area to obtain update monitoring results; if the update monitoring results meet the preset update requirements, constructing an update kinematic model corresponding to the first vehicle based on the update state data; generating an update trajectory sequence corresponding to the first vehicle based on the update kinematic model and the update state data; and sending the update trajectory sequence to the first vehicle.

[0016] Furthermore, the road intersection area is updated and monitored to obtain updated monitoring results, including: monitoring the number of vehicles in the road intersection area to obtain updated quantity results, wherein the updated quantity results are used to indicate whether the number of vehicles in the road intersection area has changed; monitoring the status of the first vehicle in the road intersection area to obtain updated status results, wherein the updated status results are used to indicate whether the vehicle status of the first vehicle meets the preset status requirements; and determining the updated monitoring results based on the updated quantity results and / or the updated status results.

[0017] Furthermore, the status update monitoring of the first vehicle in the road intersection area is carried out to obtain the status update result, including: acquiring the updated status data of the first vehicle; determining the target status deviation based on the updated status data and the target status data in the target trajectory sequence; and determining the status update result based on the target status deviation and the preset status deviation.

[0018] Furthermore, the method also includes: determining that the update monitoring result meets the preset update requirements when the quantity update result shows that the number of vehicles in the road intersection area has changed, and / or the status update result shows that the first vehicle does not meet the preset status requirements.

[0019] According to another aspect of the embodiments of this application, a trajectory generation device is also provided, applied to an edge server, comprising: an acquisition module, configured to acquire first vehicle state data of the first vehicle and second vehicle state data of at least one second vehicle located in the road intersection area when a first vehicle is detected to enter a road intersection area, wherein the road intersection area represents the area where the main road and the ramp intersect; a first construction module, configured to construct a kinematic model corresponding to the first vehicle based on the first vehicle state data, wherein the kinematic model represents the state transition relationship of the first vehicle at multiple future times; a second construction module, configured to construct a target cost function corresponding to the first vehicle based on the kinematic model, wherein the target cost function quantifies the comprehensive cost of the first vehicle in a preset time domain; a generation module, configured to generate a target trajectory sequence of the first vehicle at multiple future times based on the target cost function and the second vehicle state data; and a sending module, configured to send the target trajectory sequence to the first vehicle.

[0020] According to another aspect of the embodiments of this application, a vehicle is also provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of this application when it runs.

[0021] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0022] According to another aspect of the embodiments of this application, a computer program product is also provided, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0023] According to another aspect of the embodiments of this application, a computer program product is also provided, including a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods in various embodiments of this application.

[0024] According to another aspect of the embodiments of this application, a computer program is also provided, which, when executed by a processor, implements the methods of the various embodiments of this application.

[0025] In this embodiment, a trajectory generation method is proposed and applied to an edge server. The method, upon detecting that a first vehicle has entered a road intersection area, acquires the first vehicle's state data and the second vehicle's state data of at least one second vehicle located in the road intersection area. Then, based on the first vehicle's state data, a kinematic model representing the state transition relationships of the first vehicle at multiple future moments is constructed. Furthermore, based on the kinematic model, a target cost function corresponding to the first vehicle is constructed. And based on the target cost function and the second vehicle's state data, a target trajectory sequence of the first vehicle at multiple future moments is generated. Finally, the target trajectory sequence is sent to the first vehicle. This application employs a centralized collaborative planning approach using edge servers. By acquiring real-time state data of the first vehicle and at least one second vehicle within the intersection area when the first vehicle is detected entering the intersection, a kinematic model of the first vehicle's state transition relationships at multiple future moments is constructed. Based on this kinematic model, a target cost function for the first vehicle is built, and interaction constraints are predicted based on the target cost function and the second vehicle's state data. This generates a target trajectory sequence for the first vehicle at multiple future moments, which is then sent to the first vehicle. This achieves the goal of planning a collision-free and kinematically feasible spatiotemporal trajectory in advance, considering the dynamic interference of surrounding vehicles and the vehicle's own kinematic limitations. This eliminates single-vehicle perception blind spots, avoids game-theoretic conflicts in multi-vehicle distributed decision-making, and improves vehicle driving safety. Furthermore, it solves the technical problem of low vehicle safety when driving in intersection areas in related technologies. Attached Figure Description

[0026] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0027] Figure 1 This is a flowchart of a trajectory generation method according to an embodiment of this application;

[0028] Figure 2 This is a flowchart of a collaborative control method according to an embodiment of this application;

[0029] Figure 3 This is a flowchart of a dynamic programming solution for a spatiotemporal trajectory according to an embodiment of this application;

[0030] Figure 4 This is a diagram of an architecture for a 5G edge computing-based ramp merging collaborative control system according to an embodiment of this application.

[0031] Figure 5 This is a schematic diagram of a road intersection according to an embodiment of this application;

[0032] Figure 6 This is a schematic diagram of a trajectory generation device according to an embodiment of this application. Detailed Implementation

[0033] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0034] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0035] According to an embodiment of this application, an embodiment of a trajectory generation method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0036] This embodiment provides a trajectory generation method applied to an edge server. Figure 1 This is a flowchart of a trajectory generation method according to an embodiment of this application, such as... Figure 1 As shown, the process includes the following steps:

[0037] Step S102: When the first vehicle is detected to have entered the road intersection area, the first vehicle status data of the first vehicle and the second vehicle status data of at least one second vehicle located in the road intersection area are obtained, wherein the road intersection area is used to represent the area where the main road and the ramp intersect.

[0038] The aforementioned edge servers can refer to dedicated server clusters or high-performance computing nodes deployed at the network edge, possessing data storage, processing, and computing capabilities. The types of edge servers may include, but are not limited to, Road Side Edge Units (RSEs), Multi-access Edge Computing Servers (MECs), and roadside intelligent computing gateways; the specific edge server needs to be determined based on actual deployment requirements. Edge servers can be used as the computing entity in this application, performing data aggregation and fusion, as well as real-time collaborative trajectory planning and solving.

[0039] The aforementioned "first vehicle" can refer to any vehicle entering the road intersection area at the current moment. In merging scenarios, the first vehicle can refer to a vehicle actively merging or requiring guidance. The type of first vehicle can include vehicles merging from ramps, vehicles changing lanes on the main road, and connected and automated vehicles (CAVs), etc., with the specific first vehicle determined based on the actual situation. The first vehicle can serve as the core computational object of the trajectory generation algorithm, i.e., the planning subject. The edge server primarily uses the kinematic model of the first vehicle, combined with information from surrounding vehicles, to plan the spatiotemporal trajectory for the first vehicle, and ultimately sends this trajectory to the first vehicle for execution.

[0040] The aforementioned road intersection area refers to a pre-defined monitoring area where the main road and ramps physically intersect, overlap, or are adjacent, causing spatiotemporal conflicts in traffic flow or requiring collaborative management. The road intersection area can be covered by roadside sensing devices and centrally managed by an edge server. The road intersection area can be used to define when to start executing the trajectory generation algorithm, to define the geographical boundaries and topological constraints of trajectory generation, and to identify which vehicles (the first vehicle and the second vehicle) need to perform collaborative computation.

[0041] The aforementioned first vehicle state data can refer to a set of quantified data on the kinematic and dynamic characteristics of the first vehicle at the current moment. The first vehicle state data may include, but is not limited to, vehicle position data, vehicle speed, heading angle, acceleration, steering wheel angle, vehicle attribute data, etc., and the specific first vehicle state data needs to be determined based on the actual situation. The first vehicle state data can be used as the initial state for constructing the kinematic model of the first vehicle, determining the starting point for trajectory planning, and can be used to provide initial velocity and position, serving as a benchmark for subsequent state transitions in the time domain.

[0042] The aforementioned at least one second vehicle can refer to other vehicles located within the road intersection area that have potential spatiotemporal conflicts or interactions with the first vehicle. The type of second vehicle may include, but is not limited to, vehicles merging from ramps, vehicles changing lanes on the main road, connected and automated vehicles (CAVs), and human-driven vehicles (HDVs), etc. The specific second vehicle needs to be determined based on the actual situation. The second vehicle can be used as an environmental constraint source for the trajectory planning of the first vehicle.

[0043] The aforementioned second vehicle state data can refer to data describing the motion state of a second vehicle located within a road intersection area at the current moment and in its future predicted time domain. This second vehicle state data can be used to predict the future trajectory of the second vehicle, thereby constructing dynamic obstacle constraints; it can be used to calculate safety cost terms; and it can be used to calculate spatiotemporal constraint conditions, etc.

[0044] The aforementioned main road can refer to a lane or road segment in the traffic flow that has a large straight-through volume, relatively high speed, and priority right-of-way. The traffic density and speed distribution of the main road can be used to assess the difficulty and efficiency of merging, and the main road can be used to define the topology of road junctions.

[0045] The aforementioned ramps can refer to dedicated connecting roads of different grades or different directions of the same grade, allowing vehicles to enter or exit the main road or change their direction of travel. In merging scenarios, ramp vehicles are usually in a subordinate position, needing to find gaps to merge into or exit the main road.

[0046] In one optional embodiment, the presence of a first vehicle is determined by roadside sensing devices or by vehicle-initiated reporting. If the first vehicle is detected entering the intersection of the main road and the ramp, the first vehicle status data of the first vehicle and the second vehicle status data of at least one second vehicle within the intersection are collected. This yields a spatiotemporal state set including the target merging vehicle and surrounding related vehicles, providing a real-time and accurate data foundation for subsequent trajectory planning based on global multi-vehicle collaboration. This effectively overcomes the blind spots of single-vehicle perception and supports centralized collaborative decision-making.

[0047] In one optional embodiment, when a first vehicle is detected leaving the ramp and entering the main road and merging with the main road traffic flow, the traffic flow in the merging area is fused from multiple sources by using millimeter-wave radar and visual cameras integrated in the roadside units (RSUs) deployed around the merging area. At the same time, the first vehicle actively determines its current position, speed, and heading angle through a 5G vehicle-to-everything (V2X) communication module. The edge server synchronizes and spatially registers the second vehicle status data of at least one second vehicle (such as a vehicle adjacent to the main road lane) sensed by the roadside with the status data reported by the first vehicle, forming a unified local traffic scene dataset. This overcomes the perception limitations of single-vehicle sensors in curved or occluded environments, ensuring that the environmental information on which subsequent trajectory planning depends has global consistency and high accuracy.

[0048] Step S104: Based on the state data of the first vehicle, construct a kinematic model corresponding to the first vehicle, wherein the kinematic model is used to represent the state transition relationship of the first vehicle at multiple future moments.

[0049] The aforementioned kinematic model can refer to a set of equations describing the mathematical relationship between a vehicle's geometric position, attitude, and its motion control variables. The kinematic model can be used as the core foundation of a dynamic programming solver, used to calculate the vehicle's expected position, velocity, and heading angle at each subsequent time step (i.e., multiple future moments) based on the current vehicle state and control commands.

[0050] The aforementioned multiple future moments refer to a pre-defined set of discrete time series following the current moment. Multiple future moments can provide a search space for dynamic programming, enabling the algorithm to plan not only actions in the current second but also, proactively, the entire path for the next few seconds. By extending the prediction range of multiple future moments, it's possible to anticipate potential positional conflicts between the second and first vehicles at later points in time, thereby adjusting speeds or paths in advance to avoid sudden braking or sharp turns.

[0051] In one optional embodiment, based on the real-time motion parameters of the first vehicle in the first vehicle state data, a discrete state-space equation describing the mapping relationship between the vehicle's position, attitude, and control variables is established. This constructs a kinematic model that can characterize the longitudinal and lateral motion evolution of the first vehicle within a preset time interval, providing a mathematical basis for accurately predicting the vehicle's future trajectory, verifying path feasibility, and calculating collision costs in multi-vehicle cooperative scenarios, thus ensuring the physical rationality and computational accuracy of trajectory planning.

[0052] In one optional embodiment, based on the instantaneous speed, acceleration, and heading angle information obtained in real time from the first vehicle state data, combined with physical parameters such as vehicle wheelbase, maximum steering angle, and acceleration limits, a kinematic model is used to establish a state transition equation. Through discretization, the continuous vehicle motion is decomposed into discrete time steps with a step size of 0.1 seconds. The longitudinal displacement, lateral deviation, and heading angle change of the vehicle relative to the previous time step are calculated at each future time step. This constructs a mathematical model capable of predicting the complete motion trajectory state of the first vehicle within the next 5 seconds, providing an accurate state prediction benchmark for the dynamic programming algorithm and ensuring that the generated merging trajectory conforms to the vehicle's physical limits and possesses spatiotemporal continuity. The numerical values ​​in the above steps are only examples; specific values ​​need to be determined according to actual requirements.

[0053] Step S106: Based on the kinematic model, construct the target cost function corresponding to the first vehicle, wherein the target cost function is used to quantify the comprehensive cost of the first vehicle in a preset time domain.

[0054] The aforementioned objective cost function can be a data expression that quantifies the comprehensive cost generated by the first vehicle executing a specific trajectory within a preset time domain. The type of objective cost function can include, but is not limited to, linear cost functions, quadratic cost functions, piecewise linear or nonlinear cost functions, and weighted multi-objective cost functions, etc. The specific objective cost function needs to be determined according to actual needs. The objective cost function can be used as the optimization core of a dynamic programming algorithm. By numerically transforming abstract driving requirements into a computable minimization objective, the objective cost function finds a better trajectory by searching for the control sequence that minimizes the function value of the objective cost function.

[0055] The aforementioned preset time domain refers to the time span over which the target trajectory sequence is predicted forward. The preset time domain can be used to determine the solution range of the optimization problem, making the computational complexity controllable.

[0056] In one optional embodiment, based on the state transition relationship of the first vehicle at multiple future moments as represented by the kinematic model, and combined with the consideration of the vehicle's operating characteristics within a preset time domain, a target cost function is constructed to quantify the comprehensive cost of the first vehicle in that time domain. This allows for the consideration of multiple dimensions such as safety, efficiency, and comfort during trajectory generation, thereby achieving accurate quantitative evaluation and global optimization of the first vehicle's driving state.

[0057] Step S108: Based on the objective cost function and the second vehicle state data, generate a sequence of target trajectories for the first vehicle at multiple future moments.

[0058] The aforementioned target trajectory sequence can refer to a set of control commands, including the vehicle's position, speed, and heading angle at multiple future moments, obtained through dynamic programming based on the predictive capabilities of a kinematic model and considering the constraints of the second vehicle's state data. The target trajectory sequence may include, but is not limited to, globally optimal trajectory sequences, locally optimal trajectory sequences, and smooth spline trajectory sequences; the specific target trajectory sequence needs to be determined according to actual requirements. The target trajectory sequence can be used to ensure the spatiotemporal non-collision between the first and second vehicles, to provide an accurate reference path for the first vehicle, and to serve as a baseline trajectory. Subsequently, by comparing the deviation between the actual trajectory and the target trajectory sequence, a dynamic replanning mechanism is triggered to improve robustness.

[0059] In one optional embodiment, the future state evolution of the first vehicle is predicted based on the kinematic model of the first vehicle, and the spatiotemporal interaction boundary is determined based on the state data of the second vehicle. By constructing a target cost function that covers safety, efficiency and comfort, a dynamic programming algorithm is used to solve the problem in reverse recursion under the premise of satisfying preset driving constraints. The position, speed and heading angle sequence of the first vehicle at multiple future moments are calculated, thereby generating a globally optimal target trajectory sequence. This effectively eliminates the game conflict between vehicles, ensures the safety and smoothness of the merging process, and provides a high-precision execution reference benchmark for the vehicles.

[0060] Step S110: Send the target trajectory sequence to the first vehicle.

[0061] In one optional embodiment, a target trajectory sequence containing multiple future position, speed, and heading angle information is packaged and sent to the first vehicle via a 5G low-latency communication link. This ensures that the data is transmitted within an extremely short latency, enabling the first vehicle to receive and execute coordinated control commands in real time. This ensures the synchronization and consistency of trajectory planning between vehicles during the merging process, effectively improving the operational efficiency and driving safety of traffic flow.

[0062] In one alternative embodiment, the target trajectory sequence can be broadcast or multicast to the onboard terminal of the first vehicle using the V2X communication protocol via a roadside unit (RSU), or a dedicated data channel can be directly established via an edge server through a 5G network to transmit the target trajectory data containing the position, speed, and heading angle sequence to the first vehicle in a structured message format, ensuring that the first vehicle can receive and parse the trajectory information in real time for subsequent trajectory tracking control and to achieve coordinated movement with surrounding vehicles.

[0063] In this embodiment, a trajectory generation method is proposed and applied to an edge server. The method, upon detecting that a first vehicle has entered a road intersection area, acquires the first vehicle's state data and the second vehicle's state data of at least one second vehicle located in the road intersection area. Then, based on the first vehicle's state data, a kinematic model representing the state transition relationships of the first vehicle at multiple future moments is constructed. Furthermore, based on the kinematic model, a target cost function corresponding to the first vehicle is constructed. And based on the target cost function and the second vehicle's state data, a target trajectory sequence of the first vehicle at multiple future moments is generated. Finally, the target trajectory sequence is sent to the first vehicle. This application employs a centralized collaborative planning approach using edge servers. By acquiring real-time state data of the first vehicle and at least one second vehicle within the intersection area when the first vehicle is detected entering the intersection, a kinematic model of the first vehicle's state transition relationships at multiple future moments is constructed. Based on this kinematic model, a target cost function for the first vehicle is built, and interaction constraints are predicted based on the target cost function and the second vehicle's state data. This generates a target trajectory sequence for the first vehicle at multiple future moments, which is then sent to the first vehicle. This achieves the goal of planning a collision-free and kinematically feasible spatiotemporal trajectory in advance, considering the dynamic interference of surrounding vehicles and the vehicle's own kinematic limitations. This eliminates single-vehicle perception blind spots, avoids game-theoretic conflicts in multi-vehicle distributed decision-making, and improves vehicle driving safety. Furthermore, it solves the technical problem of low vehicle safety when driving in intersection areas in related technologies.

[0064] Optionally, based on the first vehicle state data, a kinematic model corresponding to the first vehicle is constructed, including: determining the state vector and control vector corresponding to the first vehicle based on the first vehicle state data, wherein the state vector is used to represent the kinematic state of the first vehicle at the current moment, and the control vector is used to represent the control input of the first vehicle at the current moment; determining the sampling time interval; and constructing the kinematic model based on the state vector, the control vector, and the sampling time interval.

[0065] The aforementioned state vector can refer to the set of mathematical variables describing the kinematic state of the first vehicle at a certain moment. The state vector serves as the basis for constructing the state space, quantifying the vehicle's current spatiotemporal position and motion trend. It is one of the input variables in the state transition equation of the dynamic programming algorithm, used to predict the vehicle's state evolution at future moments, and is the core basis for calculating trajectory deviation, safety distance, and cost function.

[0066] The aforementioned control vector can refer to the set of mathematical variables representing the actuator commands issued by the control system to the first vehicle at the current moment. The control vector can be used to represent the vehicle's active intervention in its motion state. In the kinematic model, the control vector serves as the control input to the state transition equation, determining the rate of change of the vehicle's state at the next moment. The dynamic programming algorithm optimizes the trajectory by searching for a better sequence of control vectors to minimize a preset cost function.

[0067] The aforementioned sampling time interval refers to the time difference between two adjacent sampling moments when the continuous-time vehicle motion process is discretized into a discrete-time series. Sampling time intervals can include, but are not limited to, 0.05s, 0.1s, 0.15s, 0.2s, etc., and the specific sampling time interval needs to be determined according to actual requirements. The sampling time interval can be used as the solution granularity of the dynamic programming solver.

[0068] In one optional embodiment, by parsing the first vehicle's state data, a state vector containing position, velocity, and heading angle, and a control vector containing acceleration and steering angle are extracted. A sampling time interval suitable for computational requirements is selected, and the continuous vehicle motion process is discretized into a discrete time series. A kinematic model based on the state transition equation is then constructed to quantitatively describe the changes in vehicle state with control input. This process establishes a mathematical mapping relationship, providing a standardized solution basis for subsequent dynamic programming algorithms, ensuring the accuracy and real-time performance of trajectory prediction. Simultaneously, the discretization reduces computational complexity and improves the system's operating efficiency on edge devices.

[0069] Optionally, the first vehicle state data includes at least: vehicle position data, vehicle speed, heading angle, acceleration, and steering wheel angle. Based on the first vehicle state data, the corresponding state vector and control vector of the first vehicle are determined, including: determining the state vector based on the vehicle position data, vehicle speed, and heading angle; and determining the control vector based on the acceleration and steering wheel angle.

[0070] The aforementioned vehicle position data can refer to the geometric coordinates of the first vehicle in a preset map coordinate system or a local coordinate system. Vehicle position data is a core element in determining the state vector, used to track the spatial distribution of vehicles in the merging zone in real time. It forms the basis for calculating Euclidean distances between vehicles, determining whether a safety boundary has been violated, and assessing merging progress, directly participating in the calculation of the safety cost function in dynamic programming.

[0071] The vehicle speed mentioned above can refer to the scalar velocity of the first vehicle along the direction of travel at the current moment. Vehicle speed can be used to determine the velocity component in the state vector, reflecting the intensity of the vehicle's motion.

[0072] The heading angle mentioned above can refer to the angle between the longitudinal axis of symmetry of the first vehicle and a preset reference direction (such as true north on a map or the direction of the lane centerline). The heading angle can be used to determine the attitude information in the state vector and describe the vehicle's orientation. The heading angle can be used as the basis for calculating the future direction of the vehicle's position change, and it can serve as an important indicator for determining whether the vehicle has completed a lane change and aligned with the merging channel. In the kinematic model, the rate of change of the heading angle (angular velocity) is directly related to the steering wheel angle.

[0073] The acceleration mentioned above can refer to the rate of change of the vehicle's velocity at the current moment. Acceleration is a core element of the control vector, representing the control input for the vehicle's longitudinal motion.

[0074] The aforementioned steering wheel angle can refer to the rotation angle of the front wheels or steering wheel of the first vehicle relative to the center position. The steering wheel angle is another core element of the control vector, representing the control input for the vehicle's lateral movement. Through the vehicle's steering geometry, the steering wheel angle determines the front wheel deflection angle, which in turn affects the vehicle's heading angle change rate, thus enabling lateral movement and path tracking.

[0075] In one optional embodiment, vehicle position data, vehicle speed, and heading angle are extracted to form a state vector representing the current kinematic state. Simultaneously, acceleration and steering wheel angle are extracted to form a control vector representing the current control input, thereby mapping discrete vehicle operation data into standardized mathematical vectors. This process, by clearly distinguishing between the two dimensions of state observation and control execution, achieves a structured encapsulation of the vehicle's physical state and control commands. This provides a clear state space and control space foundation for subsequent kinematic model construction and dynamic programming solutions, ensuring the logical rigor of trajectory calculation.

[0076] Optionally, based on the objective cost function and the second vehicle state data, a target trajectory sequence of the first vehicle at multiple future times is generated, including: determining the minimization of the objective cost function as the solution objective; solving the objective cost function based on the solution objective and the second vehicle state data to obtain the target trajectory sequence.

[0077] The aforementioned objective can refer to the goal that needs to be achieved during trajectory calculation. In this application, the objective is explicitly defined as minimizing the objective cost function, that is, finding a set of control vector sequences that makes the objective cost function reach a global minimum. The objective can provide a clear direction of convergence, ensuring that the solution to each subproblem obtained by recursion from the terminal time step is locally optimal, thereby guaranteeing that the final trajectory sequence is globally optimal.

[0078] In one optional embodiment, a target cost function incorporating safety, efficiency, and comfort terms is constructed using a kinematic model to quantify the overall driving cost within a preset time domain. Minimizing this target cost function is established as the solution objective. Using second vehicle state data as constraints, a dynamic programming algorithm is used to inversely solve the target cost function, thereby obtaining the target trajectory sequence that minimizes the overall cost. This process, by establishing a clear mathematical optimization model and introducing neighboring vehicle states as cooperative constraints, achieves centralized global optimization planning of multi-vehicle spatiotemporal trajectories, effectively eliminating game-theoretic conflicts between vehicles and improving merging efficiency and driving comfort while ensuring safe distances.

[0079] Optionally, based on the kinematic model, a target cost function corresponding to the first vehicle is constructed, including: based on the kinematic model, the vehicle speed at multiple future times, and the control vector of the vehicle at multiple future times, determining the cost value corresponding to at least one target cost term, wherein the at least one target cost term is used to represent a quantitative penalty index at at least one dimension during trajectory generation; and based on the weight coefficients corresponding to the at least one target cost term, the cost value corresponding to the at least one target cost term is weighted to construct the target cost function.

[0080] The aforementioned at least one target cost term can refer to a mathematical indicator that quantifies the degree to which a vehicle deviates from the ideal state or violates constraints in different dimensions during trajectory generation. This target cost term may include, but is not limited to, safety cost terms, efficiency cost terms, and comfort cost terms; the specific target cost term needs to be determined based on actual requirements. Target cost terms can be used to transform abstract driving strategies into concrete numerical penalties. Target cost terms can serve as the basic units for constructing the target cost function, each responsible for constraining the vehicle's behavior in different dimensions, ensuring that the generated trajectory achieves a balance between safety, economy, and comfort.

[0081] The aforementioned cost value can refer to the specific numerical result calculated at a specific moment or under a specific state for a particular objective cost term. The cost value is the quantitative representation of that cost term under the current vehicle motion state and control input.

[0082] The aforementioned weighting coefficients can refer to scalar parameters used to adjust the relative importance of each objective cost term in the total cost function. Weighting coefficients can be used to achieve trade-offs among multiple objectives.

[0083] In one optional embodiment, firstly, the vehicle speed and control vector at multiple future moments are predicted based on a kinematic model, and the specific cost value corresponding to at least one objective cost item such as safety, efficiency, and comfort at each moment is calculated; then, the cost values ​​of each item are weighted and summed using preset weight coefficients to construct an objective cost function that quantifies the comprehensive driving cost. The above process achieves a balance between vehicle driving safety, traffic efficiency, and ride comfort through multi-dimensional quantitative penalty indicators and weight configuration, enabling trajectory optimization to dynamically adjust its focus according to actual scenario requirements and improving the adaptability of the control strategy.

[0084] Optionally, based on the kinematic model, the vehicle speed at multiple future moments, and the vehicle's control vector at multiple future moments, the cost corresponding to at least one target cost term is determined, including: based on the kinematic model, determining the actual distance between the first vehicle and the second vehicle at multiple future moments, and constructing a safety cost term based on the distance difference between the actual distance between the vehicles and the preset safety distance; constructing an efficiency cost term based on the speed difference between the vehicle speed at multiple future moments and the target vehicle speed; and constructing a comfort cost term based on the rate of change of the control vector at multiple future moments.

[0085] The aforementioned actual distance between vehicles can refer to the real-time geometric distance between the first and second vehicles at multiple future moments. Actual distance between vehicles can include, but is not limited to, longitudinal distance, lateral distance, Euclidean distance, and relative distance, and the specific actual distance needs to be determined based on actual needs. Actual distance between vehicles can be used as a core physical quantity for assessing driving safety. It can reflect the potential collision risk between vehicles and is the basic input data for calculating safety costs. By predicting future actual distances between vehicles, potential collisions or rear-end accidents can be anticipated and avoided in advance.

[0086] The aforementioned preset safety distance refers to the minimum safe distance threshold that must be maintained between the first and second vehicles. The preset safety distance can be used as the lower limit of safety constraints to determine whether the current or future driving conditions are dangerous. When the actual distance between vehicles is less than or close to the preset safety distance, the safety cost will increase sharply, thereby generating an avoidance trajectory.

[0087] The aforementioned safety cost term can refer to the penalty value in the objective cost function used to quantify the risk of collision or near-collision due to insufficient vehicle spacing. The safety cost term can be used to ensure that the generated trajectory always stays away from the collision zone.

[0088] The target speed mentioned above refers to the expected speed a vehicle should achieve in the current road segment or specific scenario. The target speed can be used as a benchmark to assess the reasonableness of the current speed. By comparing the deviation between the current speed and the target speed, efficiency costs can be calculated, guiding vehicles back to the desired speed as quickly as possible and reducing traffic delays.

[0089] The efficiency cost term mentioned above can refer to the penalty value in the objective cost function used to quantify the time loss or reduced traffic efficiency caused by vehicles deviating from the target speed. The efficiency cost term can be used to encourage vehicles to maintain a higher speed, reduce unnecessary deceleration and stopping, thereby improving the overall traffic flow capacity.

[0090] The rate of change of the control vector mentioned above refers to the rate at which the control input changes over time. The rate of change of the control vector can be used to reflect the intensity of the driver's or controller's operation. High-frequency or large-amplitude control changes can lead to ride discomfort and vehicle wear. Therefore, reducing the rate of change helps generate a smooth, comfortable trajectory.

[0091] The aforementioned comfort cost term can refer to the penalty value in the objective cost function used to quantify the discomfort caused by excessive control input and its rate of change. The comfort cost term can be used to ensure that the generated trajectory conforms to the driving habits and riding preferences of human drivers, thereby improving the user experience of autonomous vehicles.

[0092] In one optional embodiment, a kinematic model is used to predict the actual distance between the first and second vehicles at multiple future moments, and the distance difference is calculated in conjunction with a preset safety distance to construct a safety cost term; simultaneously, an efficiency cost term is constructed based on the speed difference between the future vehicle speed and the target vehicle speed; and a comfort cost term is constructed based on the rate of change of the control vector. This process constrains driving safety, traffic efficiency, and passenger experience through multi-dimensional quantitative indicators, achieving a leap from simple obstacle avoidance to comprehensive performance improvement. It ensures that the generated trajectory avoids collision risks while also considering traffic efficiency and driving smoothness, thus improving the overall operational quality in merging scenarios.

[0093] Optionally, the target cost function is solved based on the solution objective and the second vehicle state data to obtain the target trajectory sequence, including: determining at least one constraint based on the second vehicle state data; and, under the premise of satisfying at least one constraint, recursively solving the target cost function based on the solution objective to obtain the target trajectory sequence.

[0094] The aforementioned constraint can refer to a physical, geometric, or logical restriction that must be satisfied during the solution process of the target trajectory sequence. The constraint may include, but is not limited to, safety constraints, kinematic constraints, spatiotemporal constraints, obstacle constraints, etc., and the specific constraint must be determined based on actual constraint requirements. The constraint can be used to ensure that the generated target trajectory sequence is physically realizable, spatially non-conflicting, and logically reasonable.

[0095] In one optional embodiment, at least one constraint condition, such as a safe distance, lane boundary, and kinematic limit, is derived based on the second vehicle state data. Under the premise of satisfying at least one of the above constraints, the objective cost function is solved by inverse recursion based on the objective of minimizing the objective cost function, thereby obtaining the target trajectory sequence that minimizes the overall cost. This process, by introducing hard constraints derived from the neighboring vehicle states, transforms the multi-vehicle cooperation problem into a restricted optimization problem, ensuring that the generated trajectory is physically executable and can avoid collisions. Simultaneously, by utilizing the global superiority of the inverse recursion algorithm, a balance between traffic efficiency and comfort is achieved while ensuring a safety baseline.

[0096] Optionally, based on the second vehicle state data, at least one constraint is determined, including: determining a safety constraint based on the actual distance between the first vehicle and the second vehicle at multiple future times; determining a kinematic constraint based on the performance limitation parameters of the first vehicle; determining a spatiotemporal constraint based on the geometric topology information of the road intersection area and the merging task requirements; and determining an obstacle constraint based on the second vehicle state data.

[0097] The aforementioned safety constraints can refer to mathematical limitations calculated based on the actual distance between the first and second vehicles at multiple future moments, used to prevent collisions or excessive closeness between the vehicles. Safety constraints may include, but are not limited to, Euclidean distance constraints, longitudinal spacing constraints, and lateral spacing constraints; the specific safety constraints need to be determined according to actual needs. Safety constraints can be used to eliminate collision risks from both geometric and spatiotemporal dimensions.

[0098] The aforementioned performance limiting parameters can refer to the limits that the physical structure and dynamics of the first vehicle itself can withstand. These performance limiting parameters can be used to provide physical boundaries for kinematic constraints. They can also be used to ensure that the generated target trajectory sequence is within the range that the vehicle hardware can execute, avoiding control failure or vehicle loss of control due to instructions exceeding the vehicle's capabilities, thus guaranteeing the executability of the trajectory.

[0099] The aforementioned kinematic constraints can refer to the constraints on control variables and their rates of change constructed based on the performance limit parameters of the first vehicle and combined with the vehicle's kinematic model. Kinematic constraints can be used to ensure that the generated target trajectory sequence not only does not exceed spatial limits but also conforms to the physical motion laws of the vehicle during its temporal evolution.

[0100] The aforementioned geometric topology information can refer to map data of the road environment. Geometric topology information may include, but is not limited to, lane centerlines, lane boundary lines, merging point locations, curve curvature, slope, curb locations, and the geometric shape and coordinates of obstacles (such as guardrails and bridge piers). Specific geometric topology information needs to be determined based on the actual road conditions. Geometric topology information provides a global spatial reference framework for trajectory planning. It can be used to define drivable and non-drivable areas for vehicles and provides geometric features of curves and slopes to assist in calculating lateral and longitudinal kinematic constraints, ensuring that vehicles always travel within drivable lanes.

[0101] The aforementioned merging task requirements refer to the specific logical objectives that vehicles must meet in the spatiotemporal dimensions to complete the merging action from the ramp to the main road. These requirements may include completing lane changes before reaching the merging point, being located in a specific lane on the main road after merging, and arriving at a specific location within a specific time window to avoid conflicts with vehicles on the main road. Specific merging task requirements need to be determined based on actual needs. These requirements can be used to transform the abstract concept of merging into concrete spatiotemporal constraints. They can also guide dynamic programming algorithms to prioritize the feasibility and timing of merging during the solution process, ensuring that vehicles can smoothly and safely integrate into the main road traffic flow.

[0102] The aforementioned spatiotemporal constraints refer to restrictions formed by combining spatial geometric constraints with the time dimension. These constraints ensure that the generated trajectory is not only spatially valid but also temporally synchronized. By introducing the time dimension, the planning algorithm can predict future environmental changes (such as the movement of other vehicles) and adjust its current decisions accordingly, achieving collaborative control.

[0103] The aforementioned obstacle constraints can refer to geometrical limitations constructed based on the second vehicle and other static or dynamic objects to avoid collisions. Obstacle constraints may include, but are not limited to, dynamic vehicle envelope constraints, static obstacle constraints, and multi-vehicle interaction constraints. Specific obstacle constraints need to be determined based on the actual conditions in the road intersection area. Obstacle constraints can be used to directly address collision risks in dynamic environments. By simplifying complex dynamic vehicles into easily computable geometric shapes (such as ellipses or rectangles) and having them exist as repulsive forces or hard constraints during trajectory calculation, it ensures that the first vehicle can avoid the second vehicle in real time, adapting to mixed traffic scenarios.

[0104] In one optional embodiment, safety constraints to prevent collisions are constructed by calculating the difference between the actual distance between the first vehicle and the second vehicle at multiple future times and the preset safe distance, thereby limiting the relative distance between the vehicles. Based on the performance limiting parameters of the first vehicle, such as the maximum acceleration and maximum steering angle, upper and lower limits of control variables and their rates of change are established to determine the kinematic constraints that satisfy the physical limits of the vehicles. Combining the geometric topological information such as the lane centerline and boundary lines of the road intersection area and the merging task requirement to complete the merging at a specific location and time, the spatial range and time window of vehicle travel are delineated to determine the spatiotemporal constraints. Based on the real-time position, speed, and size data of the second vehicle, i.e., the state data of the second vehicle, a dynamic safety envelope region that changes over time is generated to determine the obstacle constraints that avoid geometric interference with dynamic obstacles. The above process, by constructing safety boundaries based on relative position, kinematic boundaries based on vehicle physical characteristics, spatiotemporal boundaries based on road environment and timing, and dynamic obstacle boundaries based on neighboring vehicle status, forms a rigorous constraint system. This ensures that the final generated target trajectory sequence is physically executable, spatially reasonable, and temporally synchronized, effectively avoiding collision risks and meeting the timing requirements of the merging task.

[0105] Optionally, based on the second vehicle state data, the obstacle constraint conditions are determined, including: predicting the state of the second vehicle based on the second vehicle state data to obtain the predicted driving area of ​​the second vehicle at multiple future times; and determining the obstacle constraint conditions based on the predicted driving area.

[0106] The aforementioned state prediction refers to the process of calculating the state vector of the second vehicle at multiple future moments based on its current state data and using kinematic or dynamic models. State prediction can include, but is not limited to, constant speed model prediction, constant acceleration model prediction, bicycle model prediction, and probability distribution prediction. Constant speed model prediction assumes that the second vehicle will maintain its current speed and heading angle for a short period in the future, and is suitable for short-line-of-sight or straight-line driving scenarios. Probability distribution prediction can provide a deterministic trajectory or an uncertain distribution of states. The above state predictions are only examples; specific state predictions need to be determined based on actual circumstances. State prediction is the foundation for constructing obstacle constraints. State prediction transforms the instantaneous state of the second vehicle on the timeline into a series of future state sequences, providing data support for subsequently determining the spatial range it will occupy, i.e., predicting the driving area.

[0107] The predicted driving area mentioned above can refer to the set of spatial ranges that the second vehicle will occupy at multiple future moments. The predicted driving area can be used to provide a target shape for obstacle constraints, so that the planning algorithm of the first vehicle can intuitively determine whether its trajectory overlaps with or encroaches on the future position of the second vehicle, thereby defining impassable locations.

[0108] In one optional embodiment, based on the second vehicle's state data, the future position, speed, and heading angle of the second vehicle are extrapolated to obtain the predicted driving area of ​​the second vehicle at multiple future moments. This area is typically represented as a dynamic envelope containing the vehicle's geometric dimensions and safety margin. Subsequently, based on the geometric boundaries and spatiotemporal range of the obtained predicted driving area, obstacle constraints that prevent the first vehicle's trajectory from encroaching are constructed, transforming the predicted area into a hard constraint or high-cost penalty term in the optimization solution. The above process effectively solves the collision risk caused by the uncertainty of HDV driving behavior in mixed traffic scenarios by predicting the future space occupied by neighboring vehicles and transforming it into obstacle avoidance constraints, ensuring the safety of merging trajectories and real-time obstacle avoidance capabilities in dynamic environments.

[0109] Optionally, the method further includes: determining the merging point between the main road and the ramp based on road map data; and constructing a road intersection area based on the merging point.

[0110] The aforementioned road map data refers to a digital dataset used to describe the topology, geometric features, and semantic information of road networks. Road map data may include, but is not limited to, high-precision map data, topological map data, vector map data, and semantic map data; the specific road map data needs to be determined based on actual needs. Road map data can be used as a basic geographic reference framework for constructing road intersection areas.

[0111] The aforementioned merging point can refer to the spatial location where the main road lanes and ramp lanes geometrically intersect, and where vehicles merge from the ramp into the main road. The merging point can be the point where the centerline of the rightmost lane of the main road coincides with or intersects with the centerline of the ramp, or it can be the starting point where vehicles complete a lane change and formally enter the main road traffic flow. The merging point can be used to define the endpoint or critical conflict point of a merging task, and is a key geometric reference for calculating vehicle arrival times, determining merging timing windows, and delineating the upstream boundary of the merging influence zone.

[0112] In one optional embodiment, based on the geometric centerlines and boundary information of the main road and ramp lanes in the road map data, the spatial intersection point of the main road and ramp is identified and determined as the merging point. Subsequently, using this merging point as a reference, combined with a preset upstream extension distance (e.g., 150 meters for the main road and 100 meters for the ramp) and lane topology, an area including the rightmost lane of the main road and the ramp entrance section is delineated to construct the road intersection area. This process, by locking the merging geometric feature points and defining a clear collaborative control spatial range accordingly, provides a standardized geographical reference framework for subsequent vehicle entry detection, trajectory planning boundary constraints, and merging timing calculations, ensuring accurate matching between the control logic and the actual road structure. The values ​​in the above steps are for illustrative purposes only; specific values ​​need to be determined based on actual requirements.

[0113] Optionally, after sending the target trajectory sequence to the first vehicle, the method further includes: performing update monitoring on the road intersection area to obtain update monitoring results; if the update monitoring results meet preset update requirements, constructing an update kinematic model corresponding to the first vehicle based on the update state data; generating an update trajectory sequence corresponding to the first vehicle based on the update kinematic model and the update state data; and sending the update trajectory sequence to the first vehicle.

[0114] The aforementioned updated monitoring results can refer to the current system state snapshot or deviation assessment report obtained by the edge server after continuously receiving real-time status data uploaded by the first vehicle and other vehicles in the environment (such as the second vehicle) after the initial target trajectory sequence is distributed, and then performing fusion processing. The updated monitoring results can be used as input to trigger trajectory replanning. They can also provide an objective data basis for determining whether the initial trajectory is still valid, safe, and feasible. By comparing the predicted state with the actual state, the monitoring results can identify execution deviations, newly added obstacles, or sudden changes in traffic flow, thereby determining whether a new trajectory calculation needs to be initiated.

[0115] The aforementioned preset update requirements can refer to pre-set logical conditions or threshold standards used to determine whether to trigger the trajectory replanning algorithm. Preset update requirements may include, but are not limited to, changes in the number of vehicles in the road intersection area, or the first vehicle not meeting preset state requirements. Specific preset update requirements need to be determined based on actual needs. Preset update requirements can be used to balance computational resource consumption and control safety, avoiding frequent replanning when deviations are small or the environment is stable (leading to computational overload or command jitter), while ensuring immediate response in the event of significant deviations or dangers.

[0116] The aforementioned updated state data refers to the input data used to construct the updated kinematic model. The updated state data can include the state data of the first vehicle at the current replanning moment and the latest state data of other relevant vehicles in the environment (such as the second vehicle). The specific updated state data needs to be determined according to actual needs. The updated state data ensures that the newly planned trajectory starts from the vehicle's current real physical state, guaranteeing the continuity and feasibility of the trajectory. Simultaneously, the updated state data also provides the latest positions of other participants in the current environment, used to calculate new obstacle avoidance constraints.

[0117] The aforementioned updated kinematic model can refer to a mathematical model obtained by correcting or re-identifying the dynamic characteristics or actuator response characteristics of the first vehicle based on updated state data.

[0118] The aforementioned updated trajectory sequence can refer to a sequence of target states over a future period starting from the current moment, generated by rerunning dynamic programming or optimization algorithms based on updated kinematic models and updated state data.

[0119] In one optional embodiment, the vehicle status and environmental dynamics within the road intersection area are continuously monitored and fused in real time to generate updated monitoring results reflecting the current system status. When the results meet preset update requirements such as preset deviation thresholds, time cycles, or event triggers, the vehicle dynamic parameters are corrected using the real-time collected updated status data to construct an updated kinematic model that fits the actual working conditions. Based on the updated kinematic model and updated status data, the dynamic programming algorithm is re-run to generate an updated trajectory sequence starting from the current moment. Finally, the updated trajectory sequence is sent to the first vehicle. The above process achieves real-time adaptive replanning for execution deviations, model mismatches, and dynamic obstacles through a closed-loop feedback mechanism, effectively eliminating accumulated errors and ensuring the continuity, safety, and robustness of the merging trajectory in complex and changing environments.

[0120] Optionally, the road intersection area is updated and monitored to obtain updated monitoring results, including: monitoring the number of vehicles in the road intersection area to obtain a number update result, wherein the number update result is used to indicate whether the number of vehicles in the road intersection area has changed; monitoring the status of a first vehicle in the road intersection area to obtain a status update result, wherein the status update result is used to indicate whether the vehicle status of the first vehicle meets preset status requirements; and determining the updated monitoring result based on the number update result and / or the status update result.

[0121] The aforementioned quantity update result can refer to the total number of vehicles participating in collaborative control within the current road intersection area and the changes in their composition. The quantity update result is used to indicate whether the number of vehicles within the road intersection area has changed, thereby determining whether a trajectory update is needed.

[0122] The aforementioned state update result refers to the state assessment result obtained after evaluating the kinematic parameters of the first vehicle. The state update result can be used as a direct basis for triggering local trajectory correction or emergency obstacle avoidance. When the first vehicle deviates from the safe range due to execution errors, external interference, or its own operation, this result indicates that the system needs to intervene immediately and replan the trajectory to eliminate the deviation or avoid a collision.

[0123] The aforementioned preset state requirements refer to a pre-defined set of standards used to judge whether intervention is needed for the current state of the first vehicle. These preset state requirements may include, but are not limited to, safety distance thresholds, position deviation thresholds, and speed deviation thresholds; the specific preset state requirements need to be determined based on actual needs. Preset state requirements can be used as standards for state evaluation and filters for triggering replanning. They define control accuracy standards and safety baselines, ensuring that computational resources are only consumed when corrections are truly necessary, thus avoiding overreaction.

[0124] In one optional embodiment, by collecting environmental perception data and vehicle-reported information in real time within the road intersection area, the total number of vehicles in the intersection area is counted and compared with historical data to generate a quantity update result indicating the increase or decrease of the vehicle set. Simultaneously, the real-time kinematic parameters of the first vehicle, such as its pose, speed, and acceleration, are continuously acquired and compared in real time with preset deviation thresholds, safety boundaries, and driving regulations to generate a state update result reflecting whether the first vehicle's state deviates from expectations or enters a dangerous area. Combining the quantity update result and the state update result, a logical judgment is made to determine whether a replanning mechanism is triggered, thereby determining the final update monitoring result. This process, through multi-dimensional real-time monitoring and logical judgment, achieves the realization of triggering computationally intensive trajectory replanning only when actually needed, effectively reducing the system's redundant computational burden and improving the real-time response efficiency and resource utilization of the control system.

[0125] Optionally, the state update monitoring of the first vehicle in the road intersection area is performed to obtain the state update result, including: acquiring the updated state data of the first vehicle; determining the target state deviation based on the updated state data and the target state data in the target trajectory sequence; and determining the state update result based on the target state deviation and the preset state deviation.

[0126] The aforementioned target state data can refer to the set of ideal kinematic states that the first vehicle should reach at multiple future moments in the target trajectory sequence. The target state data serves as a reference trajectory for measuring the performance of the first vehicle. It provides a benchmark for state monitoring, enabling the quantification of the gap between the current actual state and the expected plan, thus determining whether the vehicle has deviated from the predetermined path or driving specifications.

[0127] The aforementioned target state deviation refers to the difference between the actual state data of the first vehicle at the current moment and the corresponding value of the target state data at the same moment. The target state deviation can be used to transform the abstract state difference into a concrete numerical value, directly used to determine whether a replanning mechanism has been triggered. The magnitude and direction of the deviation (e.g., lagging, leading, leftward, rightward) provide a basis for subsequent correction strategies.

[0128] The aforementioned preset state deviation can refer to a pre-set threshold or range used to determine whether intervention is needed for the target state deviation. Only when the actual target state deviation exceeds the preset state deviation is the state update result considered to fail to meet the preset state requirements, thereby triggering update monitoring.

[0129] In one optional embodiment, the updated state data of the first vehicle, including its position, speed, acceleration, and heading angle, are acquired in real time. This updated state data is then compared element-by-element with the target state data at the corresponding time step in the target trajectory sequence to calculate the target state deviation, including lateral position error, longitudinal position error, speed deviation, and heading angle deviation. The calculated target state deviation is then logically compared with a pre-set preset state deviation. If the target state deviation exceeds the preset state deviation, it is determined that the preset state requirement is not met, thereby generating a state update result indicating the need for replanning. This process, by quantifying the difference between the actual driving state and the ideal trajectory and comparing it with a safety boundary, achieves real-time detection of the vehicle's execution accuracy. It ensures that the subsequent replanning process is triggered only at critical points where the deviation affects safety or efficiency, effectively avoiding invalid calculations caused by minor fluctuations and improving the system's real-time performance and stability.

[0130] Optionally, the method further includes: determining that the update monitoring result meets the preset update requirements when the quantity update result shows that the number of vehicles in the road intersection area has changed, and / or the status update result shows that the first vehicle does not meet the preset status requirements.

[0131] In one optional embodiment, when changes in the number of vehicles within the road intersection area are detected, or when the current state of the first vehicle deviates from a preset state (i.e., does not meet the preset state requirements), the logic determines that the current environment or vehicle state has changed significantly, and the matching degree and safety of the original target trajectory sequence are insufficient. This determines that the updated monitoring results meet the preset update requirements. This process, through a multi-condition triggering mechanism, ensures that replanning is initiated immediately when the traffic flow topology changes or the single-vehicle execution deviation accumulates to a dangerous level. This effectively avoids control failures caused by information lag and ensures the real-time adaptability and safety of the merging process in a dynamic environment.

[0132] In one alternative embodiment, Figure 2This is a flowchart of a collaborative control method according to an embodiment of this application, such as... Figure 2 As shown, the method includes: step S201, region detection and definition; step S202, dynamic programming spatiotemporal trajectory optimization; step S203, target trajectory sequence distribution and local tracking; step S204, deviation detection; step S205, whether the deviation exceeds the limit; and step S206, merging completion and exit management. After exiting management, it re-enters step S201 to achieve closed-loop feedback.

[0133] Specifically, in step S201, area detection and definition, when a vehicle enters the merging influence zone of a highway ramp, the roadside sensing device detects the vehicle entry event, and the edge server defines the geometric boundary of the merging zone and the topological connection relationship between the main road and ramp lanes based on map data, and identifies the set of vehicles participating in the coordination, including the main road CAV, ramp CAV, and HDV as dynamic obstacles.

[0134] Step S202: Dynamic programming for spatiotemporal trajectory optimization. The edge server calls the dynamic programming solver, first defining the state vector. ,in, and Let be the vehicle's horizontal and vertical positions at time t. For vehicle speed, This is the heading angle. Next, the control vector is defined. , where a(t) is the acceleration and δ(t) is the steering wheel angle.

[0135] A state transition equation, or dynamic model, based on the vehicle kinematics model is established, as follows;

[0136] ;

[0137] in, For the transformation matrix, x is the sampling time. t+1 Let x be the state vector at the next time step. t Let u be the state vector at the current moment. t This is the control vector at the current moment.

[0138] Subsequently, a target cost function J, which includes safety costs, efficiency costs, and comfort costs, is constructed. The formula for the target cost function is as follows:

[0139] ;

[0140] Where N is the prediction time domain (e.g., 50 steps, i.e., 5-second prediction). , , These are the weighting coefficients. This is a safety cost item (penalty item, which incurs a huge cost when the distance between vehicles is less than the safe distance). This is an efficiency cost (encouraging vehicles to maintain the target cruising speed). This is a cost to comfort (penalty for sudden acceleration / deceleration and high-frequency steering). Let this be the state vector at the current moment. The control vector at the current moment, The speed is the vehicle speed.

[0141] Next, constraints such as safety constraints, kinematic constraints, and spatiotemporal constraints are set. Among them, the safety constraint requires that the Euclidean distance between any two vehicles must be greater than a preset threshold d. min The kinematic constraints are that the acceleration and steering wheel angle do not exceed the corresponding vehicle performance limits; the spatiotemporal constraints are that ramp vehicles must complete the merging action before the merging point, and the position after merging must be within a reasonable range of the main road lane.

[0142] Finally, the Bellman optimality principle is used to solve the problem in reverse from the terminal time to the initial time, calculate the optimal control quantity at each time, and generate a smooth and collision-free spatiotemporal trajectory.

[0143] In step S203, the target trajectory sequence is sent and tracked locally. The edge server sends the calculated optimal trajectory sequence to the vehicle through a 5G Ultra-Reliable and Low-Latency Communications (URLLC) link. After receiving the instruction, the vehicle controller uses model predictive control or proportional-integral-derivative (PID) algorithm to track the trajectory and drive the vehicle to travel according to the planned trajectory.

[0144] Step S204, Deviation Detection: During the execution process, the edge server continuously receives the actual location and speed information uploaded by the vehicle and calculates the deviation between the actual state and the planned state in real time.

[0145] In step S205, the system determines whether the calculated deviation exceeds the preset threshold or detects a new sudden obstacle. If the deviation exceeds the limit, the current trajectory is deemed invalid, triggering a dynamic replanning mechanism to regenerate the collision avoidance trajectory. If the deviation does not exceed the limit, the current trajectory continues to be executed.

[0146] Step S206, Merging Completion and Exit Management: When a vehicle passes through the merging influence zone (i.e., its position exceeds the preset distance downstream of the merging point), the edge server terminates the cooperative control of the vehicle, and the vehicle switches back to the normal adaptive cruise control mode. Subsequently, the system re-enters step S201 to realize continuous monitoring and closed-loop feedback control of the road intersection area until all cooperative vehicles have completed merging.

[0147] Figure 3 This is a flowchart of a dynamic programming solution for spatiotemporal trajectory according to an embodiment of this application, such as... Figure 3 As shown, the method includes: step S301, obtaining vehicle dynamics constraints; step S302, constructing state transition equations; step S303, defining cost function; step S304, solving in reverse recursion; and step S305, outputting the target trajectory sequence.

[0148] Specifically, firstly, vehicle dynamics constraints are obtained, including physical limits such as maximum acceleration and maximum steering angle, to ensure that the generated trajectory is within the vehicle's executable range. Then, a state transition equation is constructed, using a vehicle kinematic model to describe the evolution of vehicle states (such as position, velocity, and heading angle) over time, establishing a mathematical mapping relationship between the current state and the state at the next moment. Next, a cost function is defined, comprehensively considering multiple dimensions such as safety (e.g., collision avoidance penalty), efficiency (e.g., desired speed tracking), and comfort (e.g., acceleration / deceleration penalty), forming an objective function to evaluate the quality of the trajectory. Then, a reverse recursive solution strategy is adopted, based on the Bellman optimality principle, recursively calculating the optimal control quantity and corresponding minimum cumulative cost at each moment, starting from the terminal moment in the prediction time domain, thereby determining the globally optimal control sequence. Finally, the target trajectory sequence containing the ideal position, velocity, and control commands at each time step is output. The above process, through dynamic programming algorithm, provides a trajectory planning scheme with clear physical meaning and global optimization while ensuring real-time computation. Compared with black-box reinforcement learning methods, it improves the interpretability of the algorithm, facilitates functional safety certification, and ensures the smoothness and safety of the merging process.

[0149] Figure 4 This is a diagram of an architecture for a 5G edge computing-based ramp merging collaborative control system according to an embodiment of this application, such as... Figure 4 As shown, the system architecture mainly includes the vehicle layer, the edge layer, and the cloud layer.

[0150] The vehicle layer comprises at least one main road vehicle (CAV), at least one ramp vehicle (CAV), and mixed-traffic driverless vehicles (HDVs). Each CAV is equipped with onboard sensors, including but not limited to millimeter-wave radar, lidar, cameras, high-precision positioning modules, and 5G-V2X communication modules. The 5G-V2X communication module supports the 5G URLLC protocol, ensuring end-to-end latency of less than 30ms for data upload and command issuance.

[0151] The edge layer consists of edge servers deployed on the base station side. These edge servers establish a wireless connection with the vehicle layer via the 5G base station. Each edge server contains a perception fusion unit and a dynamic programming solver. The edge servers possess strong computing power capable of handling high-concurrency vehicle status data.

[0152] The cloud layer consists of a macro-level traffic model and a cloud platform, which is responsible for storing historical traffic data, performing large-scale traffic flow simulations, and updating the dynamic programming model parameters in the edge servers.

[0153] This layered architecture achieves high-precision real-time data acquisition through multi-source perception and low-latency communication at the vehicle layer, and solves the control lag problem caused by excessive latency in the cloud by utilizing the localized computing capabilities of the edge layer, ensuring millisecond-level trajectory planning and distribution in merging scenarios. At the same time, it leverages the big data accumulation and model iteration optimization capabilities of the cloud layer to form an efficient closed loop of end-edge-cloud linkage. This not only meets the requirements of 5G URLLC for low latency and high reliability, but also takes into account the real-time performance, scalability, and continuous evolution capability of the system computing and algorithm model, thereby improving the collaborative safety and traffic efficiency of the ramp merging area.

[0154] Figure 5 This is a schematic diagram of a road intersection scene according to an embodiment of this application, such as... Figure 5 As shown in the schematic diagram, 502 is the main road vehicle, 504 is the road intersection area, 506 is the main road, 508 is the manually driven vehicle, 510 is the ramp vehicle, 512 is the merging point, 514 is the ramp, and 516 is the spatiotemporal trajectory curve.

[0155] This scenario diagram intuitively illustrates how, in the high-risk and complex environment of a road intersection, the edge server generates spatiotemporal trajectory curves 516 for main road vehicle 502, ramp vehicle 510, and manually driven vehicle 508. Through centralized global optimization, it eliminates multi-vehicle game conflicts, achieving safe and efficient collaborative merging of CAV and HDV mixed traffic. This effectively solves the problems of low efficiency and safety hazards caused by blind spots in traditional single-vehicle intelligent perception and distributed decision-making, improving traffic efficiency and driving safety during the merging process.

[0156] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0157] According to an embodiment of this application, an embodiment of a trajectory generation device is provided. It should be noted that the device can be used to execute the above-described trajectory generation method.

[0158] Figure 6 This is a schematic diagram of a trajectory generation device according to an embodiment of this application, applied to an edge server, such as... Figure 6 As shown, the device includes: an acquisition module 602, a first construction module 604, a second construction module 606, a generation module 608, and a sending module 610.

[0159] The acquisition module 602 is used to acquire the first vehicle state data of the first vehicle and the second vehicle state data of at least one second vehicle located in the road intersection area when the first vehicle is detected to have entered the road intersection area. The road intersection area is used to represent the area where the main road and the ramp intersect. The first construction module 604 is used to construct a kinematic model corresponding to the first vehicle based on the first vehicle state data. The kinematic model is used to represent the state transition relationship of the first vehicle at multiple future time points. The second construction module 606 is used to construct a target cost function corresponding to the first vehicle based on the kinematic model. The target cost function is used to quantify the comprehensive cost of the first vehicle in a preset time domain. The generation module 608 is used to generate a target trajectory sequence of the first vehicle at multiple future time points based on the target cost function and the second vehicle state data. The sending module 610 is used to send the target trajectory sequence to the first vehicle.

[0160] Optionally, the construction module is used to determine the state vector and control vector corresponding to the first vehicle based on the first vehicle state data, wherein the state vector is used to represent the kinematic state of the first vehicle at the current moment, and the control vector is used to represent the control input of the first vehicle at the current moment; determine the sampling time interval; and construct a kinematic model based on the state vector, control vector and sampling time interval.

[0161] Optionally, the first vehicle state data includes at least: vehicle position data, vehicle speed, heading angle, acceleration, and steering wheel angle. The construction module is also used to determine the state vector based on the vehicle position data, vehicle speed, and heading angle; and to determine the control vector based on the acceleration and steering wheel angle.

[0162] Optionally, the generation module is used to determine the minimization of the target cost function as the solution objective; and to solve the target cost function based on the solution objective and the second vehicle state data to obtain the target trajectory sequence.

[0163] Optionally, the generation module is also used to determine the cost value corresponding to at least one target cost term based on the kinematic model, the vehicle speed at multiple future times, and the control vector of the vehicle at multiple future times, wherein the at least one target cost term is used to represent the quantization penalty index at at least one dimension during the trajectory generation process; and to construct the target cost function by weighting the cost value corresponding to the at least one target cost term based on the weight coefficients corresponding to the at least one target cost term.

[0164] Optionally, the generation module is also used to determine the actual distance between the first vehicle and the second vehicle at multiple future times based on the kinematic model, and to construct a safety cost term based on the distance difference between the actual distance between the vehicles and the preset safety distance; to construct an efficiency cost term based on the speed difference between the vehicle speed and the target vehicle speed at multiple future times; and to construct a comfort cost term based on the rate of change of the control vector at multiple future times.

[0165] Optionally, the generation module is also used to determine at least one constraint based on the second vehicle state data; and under the premise of satisfying at least one constraint, to perform a reverse recursive solution on the target cost function based on the solution objective to obtain the target trajectory sequence.

[0166] Optionally, the generation module is also used to determine safety constraints based on the actual distance between the first vehicle and the second vehicle at multiple future times; determine kinematic constraints based on the performance limitation parameters of the first vehicle; determine spatiotemporal constraints based on the geometric topology information of the road intersection area and the merging task requirements; and determine obstacle constraints based on the state data of the second vehicle.

[0167] Optionally, the generation module is also used to predict the state of the second vehicle based on the second vehicle state data, to obtain the predicted driving area of ​​the second vehicle at multiple future times; and to determine obstacle constraints based on the predicted driving area.

[0168] Optionally, the device is also used to determine the merging point between the main road and the ramp based on road map data; and to construct the road intersection area based on the merging point.

[0169] Optionally, after sending the target trajectory sequence to the first vehicle, the device is also used to update the monitoring of the road intersection area and obtain the update monitoring results; if the update monitoring results meet the preset update requirements, an update kinematic model corresponding to the first vehicle is constructed based on the update state data; an update trajectory sequence corresponding to the first vehicle is generated based on the update kinematic model and the update state data; and the update trajectory sequence is sent to the first vehicle.

[0170] Optionally, the device is further configured to monitor the number of vehicles in the road intersection area and obtain a number update result, wherein the number update result is used to indicate whether the number of vehicles in the road intersection area has changed; monitor the status update of a first vehicle in the road intersection area and obtain a status update result, wherein the status update result is used to indicate whether the vehicle status of the first vehicle meets a preset status requirement; and determine an update monitoring result based on the number update result and / or the status update result.

[0171] Optionally, the device is further configured to acquire updated status data of the first vehicle; determine the target status deviation based on the updated status data and the target status data in the target trajectory sequence; and determine the status update result based on the target status deviation and the preset status deviation.

[0172] Optionally, the device is also used to determine that the update monitoring result meets the preset update requirements when the quantity update result shows that the number of vehicles in the road intersection area has changed, and / or the status update result shows that the first vehicle does not meet the preset status requirements.

[0173] Embodiments of this application also provide a vehicle, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods described in various embodiments of this application when it runs.

[0174] Embodiments of this application also provide a computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0175] Embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0176] Embodiments of this application also provide a computer program product, including a non-volatile computer-readable storage medium for storing a computer program that, when executed by a processor, implements the methods in various embodiments of this application.

[0177] Embodiments of this application also provide a computer program that, when executed by a processor, implements the methods described in the various embodiments of this application.

[0178] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0179] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0180] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0181] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0182] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0183] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A trajectory generation method, characterized in that, Applications to edge servers include: Upon detecting that a first vehicle has entered a road intersection area, the system acquires first vehicle status data of the first vehicle and second vehicle status data of at least one second vehicle located in the road intersection area, wherein the road intersection area is used to represent the area where the main road and the ramp intersect. Based on the first vehicle state data, a kinematic model corresponding to the first vehicle is constructed, wherein the kinematic model is used to represent the state transition relationship of the first vehicle at multiple future times; Based on the kinematic model, a target cost function corresponding to the first vehicle is constructed, wherein the target cost function is used to quantify the comprehensive cost of the first vehicle in a preset time domain; Based on the target cost function and the second vehicle state data, a target trajectory sequence of the first vehicle at the multiple future times is generated; The target trajectory sequence is sent to the first vehicle.

2. The method according to claim 1, characterized in that, Based on the first vehicle state data, a kinematic model corresponding to the first vehicle is constructed, including: Based on the first vehicle state data, a state vector and a control vector corresponding to the first vehicle are determined, wherein the state vector is used to represent the kinematic state of the first vehicle at the current moment, and the control vector is used to represent the control input of the first vehicle at the current moment. Determine the sampling time interval; The kinematic model is constructed based on the state vector, the control vector, and the sampling time interval.

3. The method according to claim 2, characterized in that, The first vehicle state data includes at least: vehicle position data, vehicle speed, heading angle, acceleration, and steering wheel angle. Based on the first vehicle state data, the corresponding state vector and control vector of the first vehicle are determined, including: The state vector is determined based on the vehicle position data, the vehicle speed, and the heading angle; The control vector is determined based on the acceleration and the steering wheel angle.

4. The method according to claim 1, characterized in that, Based on the target cost function and the second vehicle state data, a target trajectory sequence of the first vehicle at the multiple future times is generated, including: Minimizing the objective cost function is determined as the solution objective; The target cost function is solved based on the target and the second vehicle state data to obtain the target trajectory sequence.

5. The method according to claim 1, characterized in that, Based on the kinematic model, a target cost function corresponding to the first vehicle is constructed, including: Based on the kinematic model, the vehicle speed at the multiple future times, and the control vector of the vehicle at the multiple future times, the cost value corresponding to at least one target cost term is determined, wherein the at least one target cost term is used to represent a quantitative penalty index in at least one dimension during trajectory generation. Based on the weight coefficients corresponding to the at least one target cost item, the cost values ​​corresponding to the at least one target cost item are weighted to construct the target cost function.

6. The method according to claim 5, characterized in that, Based on the kinematic model, the vehicle speed at the plurality of future times, and the vehicle's control vector at the plurality of future times, determine the cost value corresponding to at least one target cost term, including: Based on the kinematic model, the actual distance between the first vehicle and the second vehicle in the workshop at the multiple future times is determined, and a safety cost term is constructed based on the distance difference between the actual distance in the workshop and the preset safety distance. Based on the speed difference between the vehicle speed at the multiple future moments and the target vehicle speed, an efficiency cost term is constructed; Based on the rate of change of the control vectors at the multiple future moments, a comfort cost term is constructed.

7. The method according to claim 4, characterized in that, The target cost function is solved based on the objective and the second vehicle state data to obtain the target trajectory sequence, including: Based on the second vehicle status data, at least one constraint condition is determined; Under the premise of satisfying at least one of the constraints, the target cost function is solved in reverse recursion based on the solution objective to obtain the target trajectory sequence.

8. The method according to claim 7, characterized in that, Based on the second vehicle state data, at least one constraint condition is determined, including: Based on the actual distance between the first vehicle and the second vehicle in the workshop at the multiple future times, safety constraints are determined. Based on the performance limitation parameters of the first vehicle, determine the kinematic constraints. Based on the geometric topology information of the road intersection area and the merging task requirements, the spatiotemporal constraints are determined. Based on the second vehicle state data, obstacle constraints are determined.

9. The method according to claim 8, characterized in that, Based on the second vehicle state data, obstacle constraints are determined, including: Based on the second vehicle state data, the state of the second vehicle is predicted to obtain the predicted driving area of ​​the second vehicle at the multiple future times. The obstacle constraints are determined based on the predicted driving area.

10. The method according to claim 1, characterized in that, The method further includes: The merging point between the main road and the ramp is determined based on road map data; Based on the merging point, the road intersection area is constructed.

11. The method according to claim 1, characterized in that, After sending the target trajectory sequence to the first vehicle, the method further includes: Update monitoring was performed on the aforementioned road intersection area to obtain updated monitoring results; If the updated monitoring results meet the preset update requirements, an updated kinematic model corresponding to the first vehicle is constructed based on the updated state data. Based on the updated kinematic model and the updated state data, an updated trajectory sequence corresponding to the first vehicle is generated; The updated trajectory sequence is sent to the first vehicle.

12. The method according to claim 11, characterized in that, Updated monitoring was performed on the aforementioned road intersection area to obtain updated monitoring results, including: The number of vehicles in the road intersection area is monitored to obtain a number update result, wherein the number update result is used to indicate whether the number of vehicles in the road intersection area has changed; The first vehicle in the road intersection area is monitored for status updates to obtain status update results, wherein the status update results are used to indicate whether the vehicle status of the first vehicle meets the preset status requirements. The update monitoring result is determined based on the quantity update result and / or the status update result.

13. The method according to claim 12, characterized in that, The first vehicle in the road intersection area is monitored for status updates, and the status update results are obtained, including: Obtain the updated status data of the first vehicle; Based on the updated state data and the target state data in the target trajectory sequence, the target state deviation is determined; The state update result is determined based on the target state deviation and the preset state deviation.

14. The method according to claim 12, characterized in that, The method further includes: If the quantity update result indicates a change in the number of vehicles in the road intersection area, and / or if the status update result indicates that the first vehicle does not meet the preset status requirements, then the update monitoring result is determined to meet the preset update requirements.

15. A trajectory generation device, characterized in that, Applications to edge servers include: The acquisition module is used to acquire first vehicle status data of the first vehicle and second vehicle status data of at least one second vehicle located in the road intersection area when the first vehicle is detected to enter the road intersection area, wherein the road intersection area is used to represent the area where the main road and the ramp intersect; The first construction module is used to construct a kinematic model corresponding to the first vehicle based on the first vehicle state data, wherein the kinematic model is used to represent the state transition relationship of the first vehicle at multiple future times; The second construction module is used to construct a target cost function corresponding to the first vehicle based on the kinematic model, wherein the target cost function is used to quantify the comprehensive cost of the first vehicle in a preset time domain; The generation module is used to generate a target trajectory sequence of the first vehicle at the multiple future times based on the target cost function and the second vehicle state data; The sending module is used to send the target trajectory sequence to the first vehicle.

16. A vehicle, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the method according to any one of claims 1 to 14.

17. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device on which the storage medium is located to perform the method according to any one of claims 1 to 14.

18. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 14.

Images