A lane-changing control method and system for a vehicle based on structured risk prediction

Abstract

This invention discloses a vehicle lane-changing control method and system based on structured risk prediction, relating to the field of intelligent driving control technology. The method includes the following steps: acquiring vehicle operating status and external environment information; constructing a risk prediction model based on lane-changing intention probability, vehicle interaction graph, and dynamic response; using the risk prediction model to predict comprehensive risk based on vehicle operating status and external environment information, combined with lane-changing intention probability and vehicle interaction graph, generating a structured risk sequence; constructing an objective function containing risk terms based on the structured risk sequence, embedding risk constraints in the objective function, solving the objective function through rolling optimization to obtain the optimal control sequence; and performing vehicle lane-changing control based on the optimal control sequence. This invention can achieve synergistic optimization of safety, stability, and control performance during vehicle lane-changing.

Description

Technical Field

[0001] This invention relates to the field of intelligent driving control technology, and in particular to a vehicle lane-changing control method and system based on structured risk prediction. Background Technology

[0002] In existing technologies, lane-changing control of vehicles in complex traffic environments is typically based on trajectory prediction, estimating the vehicle's future motion state to aid control decisions. However, existing trajectory prediction methods mostly model based on vehicle states or simplified environmental information, lacking characterization of lane-changing intentions and multi-vehicle interactions. This results in risk representation failing to reflect the complex coupling characteristics of the lane-changing process. Furthermore, prediction results are usually output independently, failing to be transformed into objective functions or constraint variables in control optimization. This means control strategies still primarily rely on trajectory tracking errors or stability indices, unable to dynamically adjust according to risk changes, thus limiting the ability to proactively avoid potentially high-risk states. In addition, existing methods lack a systematic mechanism for utilizing the risk evolution process within the prediction time domain, making it difficult to achieve forward-looking control decisions for future risks. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a vehicle lane-changing control method and system based on structured risk prediction. By constructing a closed-loop coupling mechanism of "risk prediction - control optimization", the risk prediction results are structurally embedded into the Model Predictive Control (MPC) framework to achieve synergistic optimization of safety, stability and control performance during the vehicle lane-changing process.

[0004] To achieve the above objectives, the present invention is implemented through the following technical solution: The first aspect of this invention provides a vehicle lane-changing control method based on structured risk prediction, comprising the following steps: Acquire vehicle operating status and external environment information; A risk prediction model is constructed based on lane change intention probability, vehicle interaction graph, and dynamic response. The risk prediction model is used to predict comprehensive risks based on vehicle operating status and external environment information, combined with lane change intention probability and vehicle interaction graph, and generate a structured risk sequence. Specifically, the lane change intention probability of the target vehicle and surrounding vehicles is calculated based on the historical state sequence of the target vehicle and surrounding vehicles, and the vehicle interaction graph is constructed based on the relative position relationship, speed relationship and behavioral coupling relationship between vehicles during the lane change process. An objective function containing risk terms is constructed based on a structured risk sequence, and risk constraints are embedded in the objective function. The optimal control sequence is obtained by solving the objective function through rolling optimization. Vehicle lane-changing control is performed based on the optimal control sequence.

[0005] Furthermore, the vehicle's operating status includes longitudinal speed, yaw rate, sideslip angle, lateral position, heading angle, longitudinal acceleration, and lateral acceleration. External environmental information includes the relative position and relative speed of the vehicle with surrounding vehicles, and lane boundary information.

[0006] Furthermore, the comprehensive risks include collision risk, lateral drift risk, yaw stability risk, and interaction risk.

[0007] Furthermore, the risk prediction model includes a lane-changing intention layer, a vehicle interaction layer, and a dynamic response layer. Specifically, the lane-changing intention layer is constructed by calculating the probability of lane-changing intentions based on historical state sequences, which is used to characterize the likelihood of the driver performing lane-changing behavior in the future. The vehicle interaction layer is formed by constructing a vehicle interaction graph through the analysis of the multi-vehicle coupling effects during lane-changing, which is used to describe the interaction relationship between the target vehicle and surrounding vehicles. Based on vehicle motion state information, the vehicle dynamic response layer is constructed to analyze the dynamic stability response of the vehicle during lane-changing.

[0008] Furthermore, the risk is transformed into an optimization objective based on the structured risk sequence, and a risk-driven objective function is constructed based on the vehicle dynamics prediction model.

[0009] Furthermore, the specific steps for embedding risk constraints into the objective function are as follows: Set risk thresholds and determine risk constraints based on those thresholds; Dynamic reconstruction of the feasible domain based on risk status enables dynamic risk-driven operation; Risk evolution constraints and system dynamic constraints are embedded in the objective function.

[0010] A second aspect of the present invention provides a vehicle lane-changing control system based on structured risk prediction, comprising: The data acquisition module is configured to acquire vehicle operating status and external environmental information; The risk prediction module is configured to build a risk prediction model based on lane change intention probability, vehicle interaction graph and dynamic response. The risk prediction model uses vehicle operating status and external environment information, combined with lane change intention probability and vehicle interaction graph to predict comprehensive risk and generate a structured risk sequence. Specifically, the vehicle lane change intention probability is calculated based on the historical state sequence of the target vehicle and surrounding vehicles, and the vehicle interaction graph is constructed based on the relative position relationship, speed relationship and behavioral coupling relationship between vehicles during the lane change process. The control optimization module is configured to construct an objective function containing risk terms based on a structured risk sequence, embed risk constraints in the objective function, and solve the objective function through rolling optimization to obtain the optimal control sequence; The execution control module is configured to perform vehicle lane-changing control based on the optimal control sequence.

[0011] A third aspect of the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and to execute steps in the vehicle lane-changing control method based on structured risk prediction as described in the first aspect of the present invention.

[0012] A fourth aspect of the present invention provides a computer device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the vehicle lane-changing control method based on structured risk prediction as described in the first aspect of the present invention.

[0013] A fifth aspect of the present invention provides a computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps of the vehicle lane-changing control method based on structured risk prediction as described in the first aspect of the present invention.

[0014] The above one or more technical solutions have the following beneficial effects: This invention discloses a vehicle lane-changing control method and system based on structured risk prediction. To address the problems of existing trajectory prediction methods, such as the disconnect between risk prediction and control decision-making, the limited form of risk information expression making it difficult to directly participate in the control optimization process, and the lack of systematic modeling of the impact of multi-vehicle interactions and lane-changing intentions in lane-changing scenarios, this invention constructs a risk prediction model that integrates lane-changing intentions, vehicle interactions, and dynamic states, achieving a structured expression of lane-changing risks. Furthermore, this invention transforms the predicted risk sequence into control optimization variables and embeds it into the model predictive control framework through risk constraints and feasible domain reconstruction, establishing a closed-loop coupling mechanism between risk prediction and control decision-making. This enables forward-looking optimal control based on the predicted time-domain risk evolution, allowing vehicles to actively suppress risks while satisfying trajectory tracking and dynamic constraints, thereby improving the safety, stability, and environmental adaptability of the lane-changing process.

[0015] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the vehicle lane-changing control method based on structured risk prediction in Embodiment 1 of the present invention; Figure 2 This is a structural diagram of the risk prediction model in Embodiment 1 of the present invention; Figure 3 This is a framework diagram of the vehicle lane-changing control system based on structured risk prediction in Embodiment 2 of the present invention; Figure 4 This is a schematic diagram of the risk embedding control structure in Embodiment 2 of the present invention. Detailed Implementation

[0018] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] In current lane-changing control of vehicles in complex traffic environments, existing trajectory prediction methods mostly employ deterministic models or modeling approaches based on simplified probability distributions, such as outputting a single optimal trajectory or using a Gaussian distribution to describe positional uncertainty. These methods typically assume that uncertainty has unimodal or symmetrical characteristics, making it difficult to characterize the complex nonlinear and multimodal distribution characteristics resulting from the combined effects of differences in driving behavior, vehicle interactions, and environmental disturbances. Furthermore, existing methods generally focus on modeling the vehicle's own motion state, lacking a systematic description of lane-changing intentions and multi-vehicle interactions. This leads to prediction results that fail to accurately reflect the true evolution range of the vehicle's future states during lane changes, resulting in biases in the control inputs constructed based on these predictions. Consequently, this affects trajectory tracking accuracy and system stability, and may even induce safety risks under complex conditions.

[0021] While existing technologies can output risk prediction results to some extent, these results are typically only used for early warning or state assessment, lacking an effective integration mechanism with control decisions. Specifically, existing methods often output risk prediction as an independent module, while the control system still primarily optimizes trajectory tracking errors, vehicle stability indicators, or empirical rules, failing to transform risk prediction results into control optimization variables or constraints and participate in the control solution process. This separation of "risk prediction—control decision" prevents risk information from substantially constraining the control strategy, making it difficult for the control system to respond promptly to risk changes, reducing its ability to proactively avoid potential hazards, and limiting the improvement of safety and robustness during lane-changing processes.

[0022] Furthermore, the lane-changing process of vehicles exhibits significant time-varying and uncertainties, with risks continuously evolving within the prediction time domain. Existing control methods largely rely on current state or short-term forecast information for decision-making, lacking a systematic mechanism for utilizing the risk evolution process. In particular, they lack the technical means to embed risk sequences into a rolling optimization control framework and apply them to the control feasible domain or optimization constraints. This makes it difficult for control strategies to achieve forward-looking optimization for future risks, resulting in lagging control response and insufficient adaptability in multi-vehicle interactions and complex traffic scenarios.

[0023] Therefore, addressing the shortcomings of existing technologies in areas such as insufficient modeling of lane-changing intentions and vehicle interactions, limited risk representation capabilities, and difficulty in incorporating risk information into control optimization, this invention proposes a vehicle lane-changing control method and system capable of achieving structured risk modeling and deeply embedding risk prediction results into the control process. This enables optimal control oriented towards risk evolution, thereby improving the safety and control performance of the vehicle lane-changing process. Specific steps are described in the embodiments.

[0024] Example 1: Embodiment 1 of the present invention provides a vehicle lane-changing control method based on structured risk prediction, such as... Figure 1 As shown, it includes the following steps: S1: Obtain vehicle operating status and external environment information.

[0025] In one specific implementation, the vehicle's operating status includes longitudinal speed, yaw rate, sideslip angle, lateral position, heading angle, longitudinal acceleration, and lateral acceleration. External environmental information includes the relative position and relative speed of the vehicle with surrounding vehicles and lane boundary information.

[0026] Specifically, it provides a unified input for risk prediction and control optimization by collecting vehicle operating status information. and external environment information .in, , . Longitudinal velocity; This refers to the yaw rate; Side slip angle; The horizontal position; For heading angle; It is longitudinal acceleration; This is lateral acceleration. Let be the relative distance between the target car and the j-th car; Let be the relative speed between the target car and the j-th car; This refers to lane boundary information. This information is acquired through onboard sensors and a perception system, and time synchronization and coordinate unification are achieved.

[0027] S2: A risk prediction model is constructed based on lane-change intention probability, vehicle interaction graph, and dynamic response. This model uses vehicle operating status and external environmental information, combined with lane-change intention probability and vehicle interaction graph, to predict comprehensive risk and generate a structured risk sequence. Specifically, the lane-change intention probability is calculated based on the historical state sequence of the target vehicle and surrounding vehicles. A vehicle interaction graph is constructed based on the relative positional relationships, speed relationships, and behavioral coupling relationships between vehicles during the lane-change process to describe the dynamic interaction effects between multiple vehicles.

[0028] In one specific implementation, such as Figure 2 As shown, this embodiment constructs a risk prediction model based on a three-layer coupling of "lane change intention - vehicle interaction - dynamic response", generates a structured risk sequence in the prediction time domain, and uses the risk sequence as the input variable for subsequent model prediction and control.

[0029] The risk prediction model includes a lane-change intention layer, a vehicle interaction layer, and a dynamic response layer. The three-layer model is constructed and interacts with data in the order of "lane-change intention layer → vehicle interaction layer → dynamic response layer". The lane-change intention layer is used to characterize behavioral risks; the vehicle interaction layer is used to characterize environmental interaction risks; and the dynamic response layer is used to characterize vehicle stability risks.

[0030] The data flow is as follows: (1) The lane change intention layer is used to analyze the future lane change behavior tendency of the target vehicle and output the lane change intention probability.

[0031] (2) The vehicle interaction layer constructs a vehicle interaction relationship graph based on the lane change intention probability and the status of surrounding vehicles to describe the coupling effect between vehicles during the lane change process.

[0032] (3) The dynamic response layer predicts the future dynamic response and risk evolution process of the vehicle based on the vehicle dynamic state and vehicle interaction results.

[0033] Finally, the output results of the above three layers are fused by the feature fusion module to obtain the structured risk sequence in the prediction time domain.

[0034] Specifically, the historical state sequence of the target vehicle and surrounding vehicles is first collected: .

[0035] in, Represents a historical state sequence; n represents the length of the historical time window; Let represent the vehicle state vector at time k.

[0036] S2.1: Construct a risk prediction model based on lane change intention probability, vehicle interaction graph, and dynamic response.

[0037] S2.1.1: Calculate the lane-changing intention probability based on the historical state sequence, and construct a lane-changing intention layer to characterize the likelihood of the driver performing a lane-changing behavior in the future.

[0038] Specifically, the historical state sequence is input into the temporal feature extraction network for feature learning: .

[0039] in, The timing characteristics of lane-changing intentions; This is a timing coding function.

[0040] In this embodiment, the temporal coding function is implemented using a Transformer network, and the temporal correlation features in the historical state are extracted through a self-attention mechanism.

[0041] Furthermore, the probability of lane-changing intention is calculated based on temporal characteristics: .

[0042] in, The probability of a lane-changing intention; lane-change represents the event of a vehicle performing a lane-changing action; It is a historical state sequence. This represents conditional probability.

[0043] Lane change intention probability is used to describe the degree to which driving behavior affects future trajectory evolution and risk changes.

[0044] It should be noted that in some other implementations, besides constructing lane-changing intentions based on probabilistic models, other lane-changing intention prediction methods can also be used, such as intention recognition methods based on Hidden Markov Models (HMMs); ​​intention estimation methods based on Bayesian inference; driving strategy inversion methods based on Inverse Reinforcement Learning (IRL); and intention discrimination methods based on rules or expert knowledge. All of these methods can output intention variables to characterize the impact of driving behavior on risk.

[0045] S2.1.2: By analyzing the multi-vehicle coupling impact during lane changing, a vehicle interaction graph is constructed to form a vehicle interaction layer, which is used to describe the interaction relationship between the target vehicle and surrounding vehicles.

[0046] Specifically, a vehicle interaction graph is constructed to represent the multi-vehicle coupling effects during lane changing: .

[0047] in: This is a vehicle interaction diagram; For the set of vehicle nodes; This is the set of interaction edges between vehicles.

[0048] The vehicle node set includes the target vehicle and surrounding vehicles: .

[0049] in, Indicates the target vehicle; Let m represent the i-th surrounding vehicle; m represents the number of surrounding vehicles.

[0050] The vehicle interaction edge is defined as: .

[0051] in: This represents the interaction relationship between the target vehicle and vehicle j; Indicates relative distance; Represents relative velocity; Indicates the relative heading angle; This represents the function that maps interaction relationships.

[0052] Furthermore, a graph neural network is used to extract features from the vehicle interaction graph: .

[0053] in: Indicates vehicle interaction characteristics; This represents the mapping function of a graph neural network.

[0054] By using graph neural networks to dynamically mine the interaction relationships of surrounding vehicles, the multi-vehicle coupling effects during lane changing can be characterized.

[0055] It should be noted that in some other implementations, besides graph structure modeling, other methods can be used to characterize vehicle interaction relationships. These include local interaction modeling methods based on neighborhood filtering; interaction weight allocation methods based on attention mechanisms; multi-vehicle behavior modeling methods based on game theory; and methods for describing inter-vehicle forces based on potential field models. All of these methods can construct inter-vehicle coupling relationships to characterize the interaction risks during lane-changing processes.

[0056] S2.1.3: Based on vehicle motion state information, a vehicle dynamics response layer is constructed to analyze the dynamic stability response of the vehicle during lane changing.

[0057] Specifically, predicting future states based on vehicle dynamics models: .

[0058] Where: A is the system state matrix; B is the control input matrix; To control the input vector.

[0059] The control input vector is defined as: .

[0060] in: Input the steering angle; For driving torque; This is the braking torque.

[0061] Furthermore, dynamic response characteristics are constructed based on the dynamic state: .

[0062] in, Characteristics of dynamic response; This is the dynamic mapping function.

[0063] Dynamic response characteristics are used to characterize a vehicle's yaw stability, sideslip stability, and lateral dynamic response capability.

[0064] It should be noted that in some other implementations, in addition to constructing a dynamic model based on state variables, other methods can also be used, such as modeling methods based on the boundary function of the stable domain; stability discrimination methods based on Lyapunov functions; characterization methods based on tire adhesion utilization; and modeling methods based on rollover index or sideslip angle constraints.

[0065] S2.2: Based on vehicle operating status and external environment information, and combined with lane change intention probability and vehicle interaction graph, a structured risk sequence is generated to predict comprehensive risks. Comprehensive risks include collision risk, lateral drift risk, yaw stability risk, and interaction risk.

[0066] In one specific implementation, a comprehensive risk function is constructed: .

[0067] in, The overall risk at time k; This refers to the risk weighting coefficient. For collision risk; Risk of lateral offset; To mitigate the risk of horizontal oscillation; This is due to the risk of interaction.

[0068] The risk weighting coefficient is dynamically adjusted according to different working conditions to satisfy: .

[0069] in, This represents the weight corresponding to the p-th type of risk.

[0070] Specifically, collision risk is defined as: .

[0071] in, The minimum relative distance between the target vehicle and surrounding vehicles.

[0072] Lateral offset risk is defined as: .

[0073] in, This is the current horizontal position; The target lane center position.

[0074] The yaw stability risk is defined as: .

[0075] in, This refers to the vehicle's sideslip angle; ω represents the yaw rate.

[0076] Interaction risk is defined as: .

[0077] in, Let j be the influence weight of the surrounding vehicles; is the distance between the target vehicle and the j-th vehicle; i represents the target vehicle, i.e., the vehicle end; j represents the number of the surrounding vehicles; m represents the total number of surrounding vehicles.

[0078] Finally, the structured risk sequence in the prediction time domain is obtained. : .

[0079] in, This represents the comprehensive risk at time k. , To predict the length of the time domain.

[0080] The risk sequence is input into the model prediction and control module. This embodiment achieves a unified multi-source representation of lane-change risk through integrated risk coupling calculation.

[0081] It should be noted that in some other implementations, the overall risk can be characterized not only by a weighted sum but also by other methods, such as a nonlinear mapping form: Risk field form: Probability distribution form: Discretized risk representations, such as risk level classifications, can also be used. All of these different forms can serve as control inputs.

[0082] S3: Construct an objective function containing risk terms based on the structured risk sequence, embed risk constraints in the objective function, and solve the objective function through rolling optimization to obtain the optimal control sequence.

[0083] S3.1: Transform the risk into an optimization objective based on the structured risk sequence, and construct a risk-driven objective function based on the vehicle dynamics prediction model.

[0084] In one specific implementation, a model predictive control objective function is constructed that includes a risk term:

[0085] in, To predict the comprehensive optimization objective function in the time domain; To predict the length of the time domain; The lateral position deviation of the vehicle at time k; The vehicle heading angle error at time k; This represents the comprehensive risk prediction value at time k. Let k be the lateral acceleration of the vehicle at time k. Let k be the lateral impact intensity of the vehicle at time k. These are the weight coefficients corresponding to each optimization term.

[0086] Among them: lateral position deviation is used to characterize trajectory tracking error; heading angle error is used to characterize vehicle attitude deviation; comprehensive risk prediction value is used to characterize the future risk state of the vehicle; lateral acceleration and lateral impact are used to constrain the lateral stability and ride comfort of the vehicle.

[0087] By directly embedding the risk prediction value into the objective function, the risk state can directly participate in the optimization solution process, realizing the transformation from "trajectory tracking driven" to "risk driven control".

[0088] S3.2: Embed risk constraints in the objective function.

[0089] In one specific implementation, this embodiment constructs safety boundary constraints by setting a fixed risk threshold to limit vehicle operating risks from exceeding a preset safety range. It also dynamically adjusts the control feasible domain based on the risk state, achieving adaptive reconstruction of risk-related constraint boundaries. This allows the controller to tighten control degrees of freedom in advance during periods of escalating risk, thereby improving the system's proactive avoidance capability against potential hazards. Furthermore, it constrains the risk evolution trend by limiting the rate of risk growth or constructing risk decay conditions to prevent the vehicle from entering a state of rapid risk growth within the prediction time domain.

[0090] The above process constructs a risk-driven closed-loop control system from three levels: risk boundary limitation, risk adaptive adjustment, and risk dynamic evolution suppression, thereby achieving a synergistic improvement in safety, stability, and forward-looking control capabilities during lane changes. The specific steps are as follows: S3.2.1: Set risk thresholds and determine risk constraints based on risk thresholds.

[0091] Specifically, risk thresholds are set based on the risk security boundary: .

[0092] in, This represents the comprehensive risk prediction value at time k. This is the risk threshold.

[0093] When the predicted risk value exceeds the risk threshold, it indicates that the vehicle is in a high-risk state. At this time, the controller will actively adjust the control input to reduce the risk.

[0094] The risk threshold can be dynamically adjusted based on road conditions, vehicle speed, traffic density, and vehicle dynamics.

[0095] S3.2.2: Dynamically reconstruct the feasible domain of risk constraints based on risk status to achieve dynamic risk-driven operation.

[0096] In one specific implementation, the control feasible domain is dynamically adjusted according to the risk status: .

[0097] in, This is the vehicle state vector; This is a risk-related feasible domain.

[0098] Furthermore, the risk-related feasible domain is defined as: .

[0099] in, This is the vehicle state constraint function; This is the risk-related boundary adjustment function.

[0100] When risks increase This tightens control constraints; when risk decreases This expands the feasible control domain.

[0101] The aforementioned dynamic reconfiguration mechanism enables control constraints to adaptively adjust as risk changes, allowing the controller to automatically increase the strength of safety constraints in high-risk scenarios.

[0102] S3.2.3: Embed risk evolution constraints and system dynamic constraints in the objective function.

[0103] In one specific implementation, to avoid rapid risk growth, a risk evolution constraint is introduced into the objective function: .

[0104] in: This is the threshold for risk growth constraints.

[0105] Risk evolution constraints are used to limit the rate of risk growth and prevent vehicles from entering sudden high-risk states within the prediction time domain.

[0106] Furthermore, vehicle dynamics system constraints are introduced. and control input vector Simultaneously, control input constraints are added: .

[0107] And vehicle state constraints: .

[0108] in, To control the set of input constraints; This is the set of vehicle state constraints.

[0109] The aforementioned system constraints can integrate risk states, vehicle dynamics, and control inputs into the optimization solution process, achieving synergistic optimization of safety, stability, and trajectory tracking performance.

[0110] It should be noted that in some other implementations, besides embedding risks into MPC through constraints and feasible domain reconstruction, other methods can also be used, such as: (1) Risk-driven weight adaptation mechanism.

[0111] In some other implementations, the objective function weights can be dynamically adjusted based on the risk status: .

[0112] in, The dynamic weights corresponding to the q-th optimization objective; This is the risk mapping function.

[0113] When the risk increases, the weight of the risk item is increased; when the risk decreases, the weight of the trajectory tracking item is increased, thereby achieving dynamic adaptive adjustment of the control objective.

[0114] (2) Risk-driven constraint adjustment mechanism.

[0115] In some implementations, risk can be mapped to constraint boundary adjustment parameters: .

[0116] in, To control the input constraint function; This is the risk-related boundary function.

[0117] By dynamically adjusting the control input constraint boundaries according to the risk status, the control feasible region can be adaptively adjusted as the risk changes.

[0118] (3) Risk-driven control mode switching mechanism.

[0119] In some implementations, when the risk exceeds a threshold, switching is performed between different control modes, including: normal control mode, safety priority control mode, and emergency avoidance control mode.

[0120] Among them, when Normal control mode is adopted; when Adopt a safety-first control mode; when An emergency evacuation and control mode is adopted. , Thresholds corresponding to different risk levels.

[0121] (4) Risk attenuation constraint mechanism: In some implementations, risk attenuation constraints may also be introduced: .

[0122] in, Let be the risk attenuation coefficient, and: .

[0123] Risk decay constraints are used to ensure that the risk decreases over the prediction time domain, thereby improving the stability of the control system.

[0124] S3.3: Solve the objective function through rolling optimization to obtain the optimal control sequence.

[0125] In one specific implementation, this embodiment employs a finite-time-domain rolling optimization strategy.

[0126] First, the optimal control input obtained at the current time. : .

[0127] Obtain the optimal control sequence in the prediction time domain. : .

[0128] After that, only the first control input corresponding to the current moment is executed. .

[0129] The vehicle status is then updated, and the next time step optimization solution is performed again.

[0130] By repeatedly performing the above process, real-time closed-loop coupled control between risk prediction and control decision-making is achieved. It should be noted that in some other implementations, besides Model Predictive Control (MPC), other control methods can be employed, such as: Nonlinear Model Predictive Control (NMPC) for handling strongly nonlinear dynamic systems; hierarchical optimization control methods where the upper layer is risk minimization decision-making and the lower layer is trajectory tracking and stability control; reinforcement learning control methods that achieve risk-driven control decisions through policy learning; robust or adaptive control methods to cope with model uncertainties and external disturbances; and safety set-based control methods that transform risk into safety constraints by constructing a safety reachable set or control barrier function (CBF).

[0131] S4: Perform lane-changing control based on the optimal control sequence.

[0132] In one specific implementation, control quantities are sent to vehicle actuators, including steering systems, drive systems, and braking systems, according to the optimal control sequence to achieve vehicle lane-changing control.

[0133] In this embodiment, the control inputs include steering, driving, and braking. In other embodiments, they may also include steering-only control, steering and braking combined control, four-wheel independent drive / braking control, active suspension control (for stability adjustment), etc.

[0134] Compared to existing vehicle lane-changing control methods, this embodiment embeds vehicle operation risk prediction results into a model predictive control framework, transforming risk information from an auxiliary judgment variable into a control optimization variable. This solves the problem of the separation between risk prediction and control decision-making in existing technologies, thereby achieving real-time response to risk changes at the control level. By introducing the risk sequence in the prediction time domain into the control optimization process, this invention can proactively adjust potentially high-risk states during the control solution stage. Compared to methods that control only based on the current state, this effectively reduces control lag caused by prediction bias and improves the timeliness and rationality of control decisions.

[0135] Regarding control performance, this embodiment introduces a risk term and sets risk constraints in the objective function, enabling the control strategy to balance safety and stability while ensuring trajectory tracking accuracy. Under typical lane-changing simulation conditions, compared to model predictive control methods without risk constraints, the peak risk level during lane changing can be reduced by approximately 20%–35%, the yaw rate fluctuation amplitude by approximately 15%–25%, and the peak lateral acceleration by approximately 10%–20%. These results demonstrate that this embodiment is significantly effective in suppressing lateral instability and reducing potential collision risks.

[0136] In terms of engineering implementation, the method proposed in this embodiment is based on a model predictive control framework and can be directly deployed in existing autonomous driving control systems without adding extra hardware. Functional expansion can be achieved simply by integrating a risk prediction module at the software level, thereby reducing system modification costs. Furthermore, since the risk prediction results participate in the control optimization process, control deviations caused by insufficient information utilization can be reduced, improving the system's adaptability and robustness in complex traffic environments.

[0137] In summary, this embodiment has significant technical effects in improving vehicle lane changing safety, enhancing the forward-looking nature of the control system, and improving control stability, and has good engineering application value.

[0138] Example 2: Embodiment 2 of the present invention provides a vehicle lane-changing control system based on structured risk prediction, such as... Figure 3 and Figure 4As shown, it includes: The data acquisition module is configured to acquire vehicle operating status and external environmental information; The risk prediction module is configured to build a risk prediction model based on lane change intention probability, vehicle interaction graph and dynamic response. The risk prediction model uses vehicle operating status and external environment information, combined with lane change intention probability and vehicle interaction graph to predict comprehensive risk and generate a structured risk sequence. Specifically, the vehicle lane change intention probability is calculated based on the historical state sequence of the target vehicle and surrounding vehicles, and the vehicle interaction graph is constructed based on the relative position relationship, speed relationship and behavioral coupling relationship between vehicles during the lane change process. The control optimization module is configured to construct an objective function containing risk terms based on a structured risk sequence, embed risk constraints in the objective function, and solve the objective function through rolling optimization to obtain the optimal control sequence; The execution control module is configured to perform vehicle lane-changing control based on the optimal control sequence.

[0139] In terms of system implementation, various deployment methods can be adopted in some other implementations: (1) Centralized control architecture.

[0140] Risk prediction and control optimization are accomplished by a single controller.

[0141] (2) Distributed control architecture.

[0142] Risk prediction and control are implemented by different control units, which exchange data through a communication module.

[0143] (3) Vehicle-cloud collaborative architecture.

[0144] Some risk prediction calculations are completed in the cloud, while control is executed on the vehicle side.

[0145] Example 3: Embodiment 3 of the present invention provides a computer-readable storage medium storing a computer program adapted for loading by a processor and executing the steps of the vehicle lane-changing control method based on structured risk prediction as described in Embodiment 1 of the present invention.

[0146] Example 4: Embodiment 4 of the present invention provides a computer device, the device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the steps of the vehicle lane-changing control method based on structured risk prediction as described in Embodiment 1 of the present invention.

[0147] Example 5: Embodiment 5 of the present invention provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps in the vehicle lane-changing control method based on structured risk prediction as described in Embodiment 1 of the present invention.

[0148] The steps and methods involved in Examples 2, 3, 4 and 5 above correspond to those in Example 1. For specific implementation methods, please refer to the relevant description section of Example 1.

[0149] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0150] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, an optical medium, or a semiconductor medium, etc.

[0151] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A vehicle lane-changing control method based on structured risk prediction, characterized in that, Includes the following steps: Acquire vehicle operating status and external environment information; A risk prediction model is constructed based on lane change intention probability, vehicle interaction graph, and dynamic response. The risk prediction model is used to predict comprehensive risks based on vehicle operating status and external environment information, combined with lane change intention probability and vehicle interaction graph, and generate a structured risk sequence. Specifically, the lane change intention probability of the target vehicle and surrounding vehicles is calculated based on the historical state sequence of the target vehicle and surrounding vehicles, and the vehicle interaction graph is constructed based on the relative position relationship, speed relationship and behavioral coupling relationship between vehicles during the lane change process. An objective function containing risk terms is constructed based on a structured risk sequence, and risk constraints are embedded in the objective function. The optimal control sequence is obtained by solving the objective function through rolling optimization. Vehicle lane-changing control is performed based on the optimal control sequence.

2. The vehicle lane-changing control method based on structured risk prediction as described in claim 1, characterized in that, Vehicle operating status includes longitudinal speed, yaw rate, sideslip angle, lateral position, heading angle, longitudinal acceleration, and lateral acceleration. External environmental information includes the relative position and relative speed of the vehicle with surrounding vehicles and lane boundary information.

3. The vehicle lane-changing control method based on structured risk prediction as described in claim 1, characterized in that, The comprehensive risks include collision risk, lateral drift risk, yaw stability risk, and interaction risk.

4. The vehicle lane-changing control method based on structured risk prediction as described in claim 1, characterized in that, The risk prediction model comprises a lane-change intention layer, a vehicle interaction layer, and a dynamic response layer. The lane-change intention layer is constructed by calculating the probability of lane-change intentions based on historical state sequences, characterizing the likelihood of a driver performing a lane-change action in the future. The vehicle interaction layer is formed by analyzing the multi-vehicle coupling effects during lane-change, describing the interaction between the target vehicle and surrounding vehicles. Finally, the vehicle dynamic response layer is constructed based on vehicle motion state information, analyzing the dynamic stability response of the vehicle during lane-change.

5. The vehicle lane-changing control method based on structured risk prediction as described in claim 1, characterized in that, The risk is transformed into an optimization objective based on the structured risk sequence, and a risk-driven objective function is constructed based on the vehicle dynamics prediction model.

6. The vehicle lane-changing control method based on structured risk prediction as described in claim 1, characterized in that, The specific steps for embedding risk constraints into the objective function are as follows: Set risk thresholds and determine risk constraints based on those thresholds; Dynamic reconstruction of the feasible domain based on risk status enables dynamic risk-driven operation; Risk evolution constraints and system dynamic constraints are embedded in the objective function.

7. A vehicle lane-changing control system based on structured risk prediction, characterized in that, include: The data acquisition module is configured to acquire vehicle operating status and external environmental information; The risk prediction module is configured to build a risk prediction model based on lane change intention probability, vehicle interaction graph and dynamic response. The risk prediction model uses vehicle operating status and external environment information, combined with lane change intention probability and vehicle interaction graph to predict comprehensive risk and generate a structured risk sequence. Specifically, the vehicle lane change intention probability is calculated based on the historical state sequence of the target vehicle and surrounding vehicles, and the vehicle interaction graph is constructed based on the relative position relationship, speed relationship and behavioral coupling relationship between vehicles during the lane change process. The control optimization module is configured to construct an objective function containing risk terms based on a structured risk sequence, embed risk constraints in the objective function, and solve the objective function through rolling optimization to obtain the optimal control sequence; The execution control module is configured to perform vehicle lane-changing control based on the optimal control sequence.

8. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the vehicle lane-changing control system based on structured risk prediction as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1-6: a vehicle lane-changing control system based on structured risk prediction.

10. A computer device, characterized in that, include: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program that, when executed by the processor, implements the vehicle lane-changing control system based on structured risk prediction as described in any one of claims 1-6.

Images