Intelligent driving steering control method, device and equipment and storage medium

Abstract

The invention discloses an intelligent driving steering control method, device and equipment and a storage medium, and the method comprises the steps: obtaining multi-source sensing data of a vehicle, carrying out the fusion processing of the multi-source sensing data, and generating environment sensing data; predicting a behavior intention of the dynamic traffic participant based on the environmental perception data; planning a local track according to the behavior intention and the global path information, and generating a steering control instruction through an adaptive model prediction control algorithm based on the local track; and when the takeover intention of the driver is detected, a target steering control instruction is generated according to the steering control instruction and driver input, and steering control is conducted on the vehicle according to the target steering control instruction. The steering control instruction is generated through multi-source sensing data fusion, behavior intention prediction and a self-adaptive model prediction control algorithm, the target steering control instruction is generated by detecting the takeover intention of the driver, and compared with the prior art, safety, comfort and reliability of intelligent driving are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent driving technology for automobiles, and in particular to an intelligent driving steering control method, device, equipment, and storage medium. Background Technology

[0002] With the rapid development of intelligent driving technology, Advanced Driver Assistance Systems (ADAS) and Automatic Driving Systems (ADS) are gradually becoming standard features in modern intelligent connected vehicles. Steering control, as the core of vehicle lateral control, directly determines the comfort, safety, and user experience of intelligent driving.

[0003] Currently, the mainstream intelligent driving steering control solutions in the industry mainly rely on the following technical solutions: First, Lane Keeping Assist (LKA) based on lane line recognition: This solution uses cameras to identify lane lines and employs control algorithms such as PID (Proportional-Integral-Derivative) to keep the vehicle centered. This solution works reasonably well when lane lines are clear and road conditions are good, but its stability and reliability drop significantly when lane lines are worn or missing, in rainy or snowy weather, or on roads with complex curvature. Second, Tracking Control based on Navigation Path: This solution combines high-precision maps and GPS positioning information to control the vehicle to travel along a predetermined trajectory. This solution is highly dependent on high-precision maps, and there is a risk of control failure when positioning signals are lost (e.g., in tunnels or urban canyons) or map data is not updated in a timely manner. Third, traditional Model Predictive Control (MPC): This method predicts the vehicle's state over a future period by establishing a vehicle dynamics model and solving for the optimal control input. However, traditional MPC models often treat other road users as static or uniformly moving obstacles, lacking the ability to predict the intentions of dynamic targets such as surrounding vehicles and pedestrians. This results in abrupt steering behavior and poor comfort in scenarios such as cut-in and congested following, and a lack of the predictive ability of an experienced driver. Furthermore, existing systems also experience coordination conflicts at the human-machine co-driving level. When the driver intends to take over the vehicle, the system is not sensitive enough to recognize this intention, leading to a conflict between the system's control torque and the driver's input torque during the handover process, creating safety hazards. Therefore, existing technologies generally suffer from limitations such as limited perception capabilities, insufficient predictive abilities, and a lack of human-like control, resulting in insufficient safety, comfort, and reliability in intelligent driving.

[0004] Therefore, there is an urgent need for an intelligent driving steering control method that can improve the safety, comfort, and reliability of intelligent driving. Summary of the Invention

[0005] The main objective of this invention is to provide an intelligent driving steering control method, device, equipment, and storage medium, aiming to solve the technical problems of insufficient safety, comfort, and reliability of existing intelligent driving steering control technologies in complex dynamic traffic environments due to their limited perception capabilities, insufficient predictive capabilities, and lack of human-like control.

[0006] To achieve the above objectives, the present invention provides an intelligent driving steering control method, the method comprising the following steps: Acquire multi-source perception data from the vehicle and fuse the multi-source perception data to generate environmental perception data; Predict the behavioral intentions of dynamic traffic participants based on the environmental perception data; Based on the stated behavioral intent and global path information, a local trajectory is planned, and based on the local trajectory, a steering control command is generated using an adaptive model predictive control algorithm. When a driver's intention to take over is detected, a target steering control command is generated based on the steering control command and the driver's input, and the vehicle is steered according to the target steering control command.

[0007] Optionally, the step of predicting the behavioral intentions of dynamic traffic participants based on the environmental perception data includes: A traffic scene graph is constructed based on the environmental perception data. Nodes in the traffic scene graph represent dynamic traffic participants, and edges in the traffic scene graph represent the spatial and interactive relationships between the dynamic traffic participants. The traffic scene map is input into a preset graph neural network model to obtain the predicted trajectory and behavioral intention probability of each dynamic traffic participant; The predicted trajectory and the probability of behavioral intention of each dynamic traffic participant are taken as the behavioral intention of each dynamic traffic participant.

[0008] Optionally, the step of planning a local trajectory based on the behavioral intent and global path information includes: A risk assessment is conducted based on the stated behavioral intent, and the driving behavior decision of the vehicle is determined based on the risk assessment results. Based on the driving behavior decisions and global path information, a preliminary trajectory is generated; The initial trajectory is smoothed and optimized using a polynomial function to generate a local trajectory that meets a preset smoothness index.

[0009] Optionally, the step of generating steering control commands based on the local trajectory using an adaptive model predictive control algorithm includes: Based on the real-time state parameters of the vehicle and the road surface adhesion coefficient, the target prediction model corresponding to the current control cycle is determined from the model set, which includes multiple vehicle dynamics models; Based on the target prediction model and the local trajectory, the steering control sequence is rolled and optimized in the prediction time domain to obtain the optimal steering control sequence. The first control variable in the optimal steering control sequence is used as the steering control command for the current control cycle.

[0010] Optionally, the step of determining the target prediction model corresponding to the current control cycle from the model set based on the real-time state parameters of the vehicle and the road adhesion coefficient includes: Obtain the real-time status parameters and road adhesion coefficient of the vehicle; The adaptation weights of each vehicle dynamics model in the model set are determined based on the real-time state parameters and the road surface adhesion coefficient. The multiple vehicle dynamics models are fused according to the adaptation weights to obtain the target prediction model corresponding to the current control cycle.

[0011] Optionally, the step of generating a target steering control command based on the steering control command and the driver input when the driver's intention to take over is detected, and then performing steering control on the vehicle based on the target steering control command, includes: Real-time monitoring of the driver's visual characteristics, head posture, and hand torque; Based on the aforementioned visual features, head posture, and hand torque, determine whether the driver has an intention to take over. When the driver intends to take over, the steering control command and the driver input are weighted and fused based on the smooth handover strategy to generate the target steering control command. The vehicle is steered according to the target steering control command.

[0012] Optionally, the step of acquiring multi-source perception data of the vehicle and fusing the multi-source perception data to generate environmental perception data includes: Acquire multi-source perception data of the vehicle, perform spatiotemporal synchronization processing on the multi-source perception data, and obtain processed multi-source perception data; The processed multi-source sensing data is fused using a deep learning model and Kalman filtering to generate environmental sensing data.

[0013] Furthermore, to achieve the above objectives, the present invention also proposes an intelligent driving steering control device, the device comprising: The data processing module is used to acquire multi-source perception data of the vehicle and perform fusion processing on the multi-source perception data to generate environmental perception data. The intent prediction module is used to predict the behavioral intent of dynamic traffic participants based on the environmental perception data. The instruction generation module is used to plan a local trajectory based on the behavioral intent and global path information, and generate steering control instructions based on the local trajectory using an adaptive model predictive control algorithm. The control handover module is used to generate a target steering control command based on the steering control command and the driver input when the driver's intention to take over is detected, and to perform steering control on the vehicle according to the target steering control command.

[0014] Furthermore, to achieve the above objectives, the present invention also proposes an intelligent driving steering control device, the device comprising: a memory, a processor, and an intelligent driving steering control program stored in the memory and executable on the processor, the intelligent driving steering control program being configured to implement the steps of the intelligent driving steering control method as described above.

[0015] Furthermore, to achieve the above objectives, the present invention also proposes a storage medium storing an intelligent driving steering control program, wherein the intelligent driving steering control program, when executed by a processor, implements the steps of the intelligent driving steering control method as described above.

[0016] This invention discloses a method for acquiring multi-source perception data from a vehicle and fusing this data to generate environmental perception data; predicting the behavioral intentions of dynamic traffic participants based on the environmental perception data; planning a local trajectory based on the behavioral intentions and global path information; and generating steering control commands based on the local trajectory using an adaptive model predictive control algorithm; and generating a target steering control command based on the steering control command and driver input when a driver takeover intention is detected, and then steering the vehicle according to the target steering control command. Because this invention fuses multi-source perception data, predicts the behavioral intentions of dynamic traffic participants, plans a local trajectory based on these intentions, generates steering control commands using an adaptive model predictive control algorithm, and finally generates a target steering control command based on the steering control command and driver input when a driver takeover intention is detected, this invention improves the safety, comfort, and reliability of intelligent driving compared to existing technologies. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the first embodiment of the intelligent driving steering control method of the present invention; Figure 2This is a schematic diagram of the overall system architecture of the intelligent driving steering control method of the present invention; Figure 3 This is a schematic diagram illustrating the principle framework of adaptive model predictive control in the intelligent driving steering control method of the present invention. Figure 4 This is a flowchart illustrating the second embodiment of the intelligent driving steering control method of the present invention; Figure 5 This is a flowchart illustrating the behavioral intent prediction process in the intelligent driving steering control method of the present invention. Figure 6 This is a flowchart illustrating the third embodiment of the intelligent driving steering control method of the present invention; Figure 7 This is a structural block diagram of the first embodiment of the intelligent driving steering control device of the present invention; Figure 8 This is a schematic diagram of the structure of an intelligent driving steering control device in the hardware operating environment involved in the embodiments of the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0020] This invention provides an intelligent driving steering control method, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the intelligent driving steering control method of the present invention.

[0021] In this embodiment, the intelligent driving steering control method includes steps S10 to S40: Step S10: Acquire multi-source perception data of the vehicle and perform fusion processing on the multi-source perception data to generate environmental perception data.

[0022] It should be noted that the executing entity in this embodiment can be a computer server device with data processing, network communication, and program execution functions applied in intelligent driving scenarios, such as the central control unit of an intelligent driving system, or an electronic device capable of realizing the above functions (such as an intelligent driving steering control device). The following uses a system including an intelligent driving steering control device (hereinafter referred to as the system) as an example to illustrate this embodiment and the following embodiments.

[0023] It should be understood that multi-source sensing data can refer to the raw or preliminary processed data set of information about the vehicle's own state and external environment collected by sensors of different types and physical principles.

[0024] In specific implementations, for example, refer to Figure 2 Multi-source sensing data can be collected through vehicle cameras, millimeter-wave radar, lidar, high-precision maps, GPS / IMU positioning units, and vehicle bus.

[0025] Understandably, environmental perception data refers to data generated after fusing multi-source perception data, which can include static environmental perception results and dynamic target tracking information. Static environmental perception results can include lane line geometric parameters (curvature, width), roadside position, drivable area boundaries, and semantic information of traffic signs. Dynamic target tracking information can include the precise position, speed, acceleration, and heading angle of surrounding vehicles, pedestrians, non-motorized vehicles, and other dynamic objects, and assign a tracking ID to each target.

[0026] It should be noted that fusing multi-source sensing data can enable complementary advantages of different sensors (e.g., improving data accuracy and reliability); it can also create redundancy (e.g., when one sensor fails or degrades due to specific reasons, other sensors can provide backup information to ensure that the system can still maintain basic environmental perception capabilities and will not immediately collapse), thus enhancing the system's robustness.

[0027] In a specific implementation, multi-source perception data of the vehicle can be acquired, and the multi-source perception data can be spatiotemporally synchronized to obtain processed multi-source perception data. The processed multi-source perception data can then be fused using a deep learning model and Kalman filtering to generate environmental perception data.

[0028] It should be understood that, since the data collected by different sensors are collected at different times and in different coordinate systems, spatiotemporal synchronization processing of multi-source sensing data can unify the information to the same time and space reference, thereby forming a consistent and coherent environmental picture.

[0029] It should be noted that the deep learning model is used to perform semantic understanding and feature extraction on the processed multi-source perceptual data, and then Kalman filtering is used for fusion processing. The deep learning model provides high-level semantic understanding, while Kalman filtering provides low-level, smooth, and physically plausible state trajectories. This ensures the accuracy and smoothness of the system.

[0030] Step S20: Predict the behavioral intentions of dynamic traffic participants based on the environmental perception data.

[0031] It should be understood that dynamic traffic participants can refer to objects in a vehicle-driving environment that possess the ability to move and whose state (such as position, speed, and direction) changes over time. Dynamic traffic participants can include other vehicles (e.g., vehicles in front, beside, or approaching from the opposite lane that may potentially affect this vehicle), pedestrians, non-motorized vehicles, and other movable objects (e.g., pets, scooters, etc.).

[0032] It's important to clarify that behavioral intent refers to a driving or movement action with clear semantics that a dynamic traffic participant intends to perform in the near future. It's not merely a prediction of their future physical trajectory, but rather an inference of their behavioral purpose. For example, for a vehicle beside you: its behavioral intent might be to maintain its current lane, potentially cut in front of you, or potentially leave your lane.

[0033] Understandably, predicting the behavioral intentions of dynamic traffic participants can overcome the lag and passivity of traditional control. For example, by predicting that a vehicle intends to cut in (even before it has started moving), the system can make a prediction tens of milliseconds or even seconds in advance. This allows the system to proactively and in advance make small speed adjustments or smooth lateral avoidance, rather than swerving at the last minute, upgrading from a passive reaction to proactive coordination.

[0034] Furthermore, predicting the behavioral intentions of dynamic traffic participants enables a more human-like and comfortable driving experience. Human drivers (especially experienced drivers) constantly observe and interpret the intentions of surrounding vehicles. For example, seeing a neighboring vehicle's wheels swerve or the driver frequently checking their rearview mirror will indicate a potential lane change, allowing them to release the accelerator and prepare to brake in advance. This embodiment, by predicting the behavioral intentions of dynamic traffic participants, enables more human-like, smooth, and comfortable decision-making and control.

[0035] In practical implementation, the behavioral intentions of dynamic traffic participants can be predicted using a behavior prediction model based on the environmental perception data. The behavior prediction model can be a graph neural network model or a long short-term memory network model, etc.

[0036] Step S30: Plan a local trajectory based on the behavioral intent and global path information, and generate steering control commands based on the local trajectory using an adaptive model predictive control algorithm.

[0037] Understandably, global path information refers to the long-term, macroscopic driving route of a vehicle from its origin to its destination. It typically originates from the vehicle's navigation system and high-precision maps, and includes information such as road level (e.g., highway, main road), approximate direction, and key nodes to be passed (e.g., intersections, ramps). Global path information ensures that the planning of local trajectories always serves the macroscopic goal of reaching the final destination.

[0038] It should be understood that a local trajectory can refer to the specific and precise path that a vehicle will travel within a short distance over the next few seconds to tens of seconds, guided by the global path. It includes the precise position (lateral and longitudinal), attitude, speed, and acceleration that the vehicle should be in at any given moment.

[0039] Planning a local trajectory based on behavioral intent and global path information enables proactive and smooth responses to dynamic environments. Without behavioral intent prediction, the system would treat surrounding vehicles as obstacles moving at a constant speed. When a vehicle begins to encroach on the lane (intending to cut in), the system might fail to recognize this in advance, only making a sudden emergency swerve after the other vehicle has already violated the lane boundary, resulting in harsh steering and poor comfort. By combining behavioral intent, the system can begin planning a local trajectory with continuous curvature and gentle acceleration changes much earlier, rather than making a sharp turn at the last minute, thus improving the comfort of intelligent driving.

[0040] In a specific implementation, a target prediction model corresponding to the current control cycle can be determined from a model set based on the real-time state parameters of the vehicle and the road surface adhesion coefficient. The model set includes multiple vehicle dynamics models. Based on the target prediction model and the local trajectory, the steering control sequence is rolled and optimized in the prediction time domain to obtain the optimal steering control sequence. The first control quantity in the optimal steering control sequence is used as the steering control command for the current control cycle.

[0041] Understandably, the real-time state parameters of a vehicle can refer to a series of key physical quantities that characterize the vehicle's motion state and attitude at the current moment, such as vehicle speed and yaw rate.

[0042] It's important to explain that traditional model predictive control (MPC) algorithms typically use a fixed vehicle dynamics model. However, in actual operation, vehicle load, speed, and the coefficient of friction between tires and the road surface (road adhesion coefficient) all change, making a fixed model inaccurate. Adaptive model predictive control (A-MPC) algorithms can automatically adjust or switch the vehicle dynamics model they use based on real-time monitored vehicle conditions (such as vehicle speed and yaw rate) and road conditions, thus maintaining high-precision prediction and control performance under various operating conditions.

[0043] It should be understood that the predictive time domain refers to the time frame within which the controller anticipates the future state of the vehicle and optimizes control parameters from the current moment. Within this future time window, the controller simulates the vehicle's behavior based on the vehicle model. By optimizing the control sequence throughout the predictive time domain, it ensures that the vehicle smoothly and stably tracks the target trajectory not only in the current moment but also over a short period in the future. This gives the system "forward-looking" capabilities and is key to improving control smoothness and human-likeness.

[0044] It should be noted that the current control cycle refers to the time interval during which the system executes one complete "sensing-computation-control" cycle. For example, 10 milliseconds (ms). At the beginning of each "current control cycle," the system acquires the latest sensor data (such as real-time state parameters), then performs all calculation steps (e.g., determining the target prediction model, rolling optimization), and finally outputs a steering control command. The next cycle repeats this process, thus ensuring the real-time performance of the control.

[0045] It needs to be explained that rolling optimization refers to solving for an optimal front wheel angle sequence (i.e., the optimal steering control sequence) in each current control cycle with the goal of tracking the local trajectory, taking into account constraints such as control error, steering wheel angle change rate, and lateral acceleration in the prediction time domain, and outputting the first control quantity to the steer-by-wire (SBW) system, which is the final actuator.

[0046] It should be noted that the optimal steering control sequence can refer to a series of future front wheel steering angle control quantities obtained by solving an optimization problem in the prediction time domain. The optimal steering control sequence can refer to the "best" future operation scheme obtained by mathematical optimization calculation after comprehensively considering various constraints (such as minimizing tracking error, steering wheel angle change rate, lateral acceleration, etc.).

[0047] It should be noted that the Adaptive Model Predictive Control (A-MPC) algorithm achieves vehicle dynamics model adaptation: it incorporates multiple vehicle dynamics models (such as linear two-degree-of-freedom models, nonlinear models, etc.), and the controller automatically selects or fuses the most suitable target prediction model based on real-time parameters such as the current vehicle speed and the estimated road adhesion coefficient, thereby significantly improving the control accuracy and robustness under different operating conditions.

[0048] In a specific implementation, the real-time state parameters and road surface adhesion coefficient of the vehicle can be obtained; the adaptation weights of each vehicle dynamics model in the model set can be determined based on the real-time state parameters and the road surface adhesion coefficient; and multiple vehicle dynamics models can be fused based on the adaptation weights to obtain the target prediction model corresponding to the current control cycle.

[0049] It should be understood that the road surface adhesion coefficient is a dimensionless parameter that characterizes the maximum static friction force between the tire and the road surface. It directly determines the maximum longitudinal (acceleration / braking) and lateral (steering) forces that the vehicle can obtain. Different road surface adhesion coefficients will greatly affect the dynamic limits of the vehicle. This embodiment estimates or senses the road surface adhesion coefficient, enabling the adaptive model predictive control algorithm to dynamically adjust its internal target prediction model, ensuring that stable control commands that do not exceed physical limits can be generated under different road conditions (dry and wet).

[0050] It should be added that, in order to improve the reliability of intelligent driving, feedback correction is also performed based on the error between the actual state and the predicted state of the vehicle to achieve closed-loop optimization.

[0051] It should be understood that the actual vehicle state refers to a set of physical quantities that describe the instantaneous motion attitude and dynamic behavior of the vehicle, actually measured or estimated by the vehicle's own sensors at the beginning of the current control cycle. The predicted state refers to the state that the vehicle should be in over a future period of time (i.e., within the prediction time domain), as calculated by the A-MPC controller based on its internal vehicle dynamics model and the control sequence calculated in the previous control cycle.

[0052] For example, refer to Figure 3 The reference state (i.e., predicted state) and ideal trajectory (i.e., local trajectory) input from the upper layer provide the control objective for the A-MPC controller. The A-MPC controller obtains the optimal steering control sequence through rolling optimization (combining cost function and dynamic constraints). The adaptive update of model parameters continuously optimizes the model through the estimated parameters of the parameter estimator, thereby determining the target prediction model corresponding to the current control cycle. The state feedback transmits the real-time state of the vehicle system back to achieve closed-loop control. Finally, the A-MPC controller outputs the first control quantity in the optimal steering control sequence and applies it to the vehicle system. The state output by the vehicle system then drives subsequent parameter estimation, model update and optimization iteration, forming a complete adaptive model predictive control process.

[0053] Step S40: When the driver's intention to take over is detected, a target steering control command is generated based on the steering control command and the driver's input, and the vehicle is steered according to the target steering control command.

[0054] It should be noted that, in order to eliminate the potential for human-machine conflict and make human-machine co-driving more harmonious and safe, the driver's visual characteristics, head posture, and hand torque can be monitored in real time. Based on the aforementioned visual characteristics, head posture, and hand torque, it can be determined whether the driver intends to take over. When the driver intends to take over, the steering control command and the driver's input are weighted and fused based on a smooth handover strategy to generate a target steering control command. The vehicle is then steered according to the target steering control command.

[0055] It should be explained that visual characteristics refer to a series of visual physiological parameters acquired by the driver monitoring system's cameras that reflect the driver's attention state and potential driving intentions, including the driver's gaze direction and eye state. Head posture refers to the angle and orientation of the driver's head. Hand torque refers to the torque applied by the driver's hands to the steering wheel.

[0056] It should be understood that the target steering control command can refer to the steering control signal generated by the system and ultimately executed when the driver intends to take over control of the vehicle. This signal integrates the system's intelligent driving command (i.e., the steering control command) and the driver's manual input command (i.e., the driver's input).

[0057] In practical implementation, the driver's visual characteristics, head posture, and hand torque can be monitored in real time using a DMS (Driver Monitoring System) camera. When the system detects that the driver intends to take over the steering wheel (e.g., the driver turns their head back to face the road and keeps their hands on the steering wheel while applying a torque exceeding a threshold T), control is smoothly transferred to the driver in a gradual manner (i.e., a smooth handover strategy). Specifically, steering control commands and driver input are weighted and fused, with weights... The system smoothly transitions from 1 (full system control) to 0 (full driver control) within an extremely short time (e.g., 0.5 seconds), avoiding sudden torque changes and ensuring a safe and natural handover. The corresponding mathematical expression is: ; In the formula, This indicates a target steering control command. This indicates a steering control command. This indicates input from the driver.

[0058] Understandably, when the driver does not intend to take over, the steering control command is taken as the target steering control command, and the vehicle is steered according to the target steering control command.

[0059] refer to Figure 2 , Figure 2This is a schematic diagram of the overall system architecture of the intelligent driving steering control method of the present invention. The system architecture is divided into three parts: the input layer / sensors, the core processing layer, and the output layer actuators. The input layer / sensors include cameras, millimeter-wave radar, lidar, high-precision maps & GPS / IMU, the DMS driver monitoring system, and the vehicle CAN bus, which provide multi-source perception data for the system. The core processing layer includes a multi-modal perception fusion module (receiving input layer data and outputting environmental perception information to the dynamic scene prediction and intent recognition module), a dynamic scene prediction and intent recognition module (outputting predicted intent and trajectory based on environmental perception information, i.e., behavioral intent, while receiving driver status from the DMS driver monitoring system, vehicle status from the vehicle CAN bus, and anthropomorphic smooth trajectory feedback from the collaborative decision-making and trajectory planning module to obtain a local trajectory), a collaborative decision-making and trajectory planning module (receiving predicted intent and trajectory from the dynamic scene prediction and intent recognition module, combining driver status and vehicle status, outputting anthropomorphic smooth trajectory, i.e., local trajectory, to the dynamic scene prediction and intent recognition module, while outputting control transfer instructions and outputting commands to the vehicle dynamics prediction and control module (A-MPC), and a vehicle dynamics prediction and control module (A- The system receives instructions from the Cooperative Decision-Making and Trajectory Planning module (MPC) and outputs steering wheel angle control instructions to the output layer actuator. The output layer actuator is the Steer-by-Wire (SBW) system, which receives and executes steering wheel angle control instructions from the Vehicle Dynamics Predictive Control module to control vehicle movement. The entire system achieves an autonomous driving process from environmental perception, intent recognition, decision-making and planning to execution control through information transmission and processing between modules.

[0060] This embodiment discloses acquiring multi-source perception data of the vehicle and fusing the multi-source perception data to generate environmental perception data; predicting the behavioral intentions of dynamic traffic participants based on the environmental perception data; planning a local trajectory based on the behavioral intentions and global path information; and generating steering control commands based on the local trajectory using an adaptive model predictive control algorithm; when a driver's intention to take over is detected, generating a target steering control command based on the steering control command and driver input, and then performing steering control on the vehicle according to the target steering control command. Because this embodiment fuses multi-source perception data, predicts the behavioral intentions of dynamic traffic participants, plans a local trajectory based on the behavioral intentions, generates steering control commands using an adaptive model predictive control algorithm, and finally generates a target steering control command based on the steering control command and driver input when a driver's intention to take over is detected, this embodiment improves the safety, comfort, and reliability of intelligent driving compared to existing technologies.

[0061] refer to Figure 4 , Figure 4This is a flowchart illustrating the second embodiment of the intelligent driving steering control method of the present invention.

[0062] Based on the first embodiment described above, in this embodiment, step S20 includes steps S201 to S203: Step S201: Construct a traffic scene graph based on the environmental perception data. The nodes in the traffic scene graph represent dynamic traffic participants, and the edges in the traffic scene graph represent the spatial and interactive relationships between the dynamic traffic participants.

[0063] Step S202: Input the traffic scene map into a preset graph neural network model to obtain the predicted trajectory and behavioral intention probability of each dynamic traffic participant.

[0064] Step S203: The predicted trajectory and the probability of behavioral intention of each dynamic traffic participant are taken as the behavioral intention of each dynamic traffic participant.

[0065] It should be understood that a traffic scene map can refer to a mathematical model used to digitally describe the entire traffic environment at a given moment. It abstracts the complex physical world into a network graph composed of "points" and "lines". Spatial relationships represent the relative positions and distances between dynamic traffic participants (e.g., vehicle A is in the left lane of vehicle B, 5 meters away), and interaction relationships can refer to the potential behavioral influences between dynamic traffic participants (e.g., a vehicle's deceleration will affect the following strategy of vehicles behind; a pedestrian stopping at an intersection may affect the decision of vehicles turning right).

[0066] It should be noted that the preset graph neural network model can refer to a fixed graph neural network model that has been trained and optimized on a large amount of historical traffic data before the system is actually deployed and run.

[0067] It should be explained that predicted trajectory refers to a quantitative prediction of the future physical movement path of dynamic traffic participants. Predicted trajectory can provide key physical constraints for trajectory planning, ensuring that the planned path of the vehicle will not collide with any predicted path. Behavioral intention probability refers to a qualitative judgment of the driving behavior that a dynamic traffic participant is about to perform, and its uncertainty is expressed in probabilistic form. Behavioral intention probability can provide high-level semantic information for decision-making and planning, and the system can make strategy adjustments in advance based on this probability value.

[0068] For example, refer to Figure 5 , Figure 5This is a schematic diagram of the behavioral intent prediction process in the intelligent driving steering control method of this invention. First, the system's multimodal perception module outputs time-series data (i.e., environmental perception data) as the basis for subsequent processing. Next, it enters the dynamic interactive scene graph construction stage, first defining nodes and edges, then generating node features and edge features respectively, completing the structure and feature initialization of the traffic scene graph. Afterwards, the traffic scene graph is input to a preset graph neural network (GNN) module, sequentially passing through a node encoder (MLP) and an edge encoder (MLP) to encode the node and edge features. Then, message passing (multi-layer MPNN) enables interactive updates of information between nodes. Combined with global context aggregation (attention mechanism), global information is further integrated, finally updating the node embedding to obtain a node representation containing rich interactive information. Subsequently, the decoder (MLP + attention mechanism) decodes the updated node embedding, outputting "predicted intent and trajectory," which is divided into two branches: one is intent classification (Softmax), outputting the prediction result (i.e., behavioral intent probability, such as "Cut - ...). The probability of an adjacent vehicle changing lanes and cutting into the lane of this vehicle is 85%; the second is trajectory prediction (multimodal probability distribution), which outputs predicted trajectories (multiple probabilistic trajectories), thereby completing the entire process from multimodal time series data to intent and trajectory prediction.

[0069] This embodiment discloses a method for constructing a traffic scene map based on environmental perception data. Nodes in the traffic scene map represent dynamic traffic participants, and edges represent the spatial and interactive relationships between these dynamic traffic participants. The traffic scene map is input into a preset graph neural network model to obtain the predicted trajectories and behavioral intention probabilities of each dynamic traffic participant. These predicted trajectories and behavioral intention probabilities are then used as the behavioral intentions of each dynamic traffic participant. Because this embodiment constructs a traffic scene map based on environmental perception data and inputs it into a preset graph neural network model to obtain the behavioral intentions of each dynamic traffic participant, compared to existing technologies, this embodiment enhances the anthropomorphism and intelligence level of intelligent driving.

[0070] refer to Figure 6 , Figure 6 This is a flowchart illustrating the third embodiment of the intelligent driving steering control method of the present invention.

[0071] Based on the above embodiments, in this embodiment, step S30 includes steps S301 to S303: Step S301: Conduct a risk assessment based on the stated behavioral intent, and determine the driving behavior decision of the vehicle based on the risk assessment results.

[0072] Step S302: Generate a preliminary trajectory based on the driving behavior decision and global path information.

[0073] Step S303: Use a polynomial function to smooth and optimize the preliminary trajectory to generate a local trajectory that meets the preset smoothness index.

[0074] It should be understood that through risk assessment and driving behavior decision-making, the system can identify potential risks in advance (such as other vehicles cutting in) and proactively decide whether to yield or accelerate through, rather than waiting for a danger to occur before braking or turning sharply. The entire decision-making and planning process mimics human driving logic (observation-judgment-decision-action), making vehicle behavior more predictable and easier for human drivers and passengers to understand and trust.

[0075] Furthermore, the smoothing optimization process and smoothness index directly address the problems of stiff steering and frequent corrections ("steering wheel effect") in traditional autonomous driving. The generated trajectory curvature changes continuously, avoiding abrupt acceleration and steering, and greatly improving comfort.

[0076] It should be noted that risk assessment can transform behavioral intentions into specific, quantifiable levels of danger, providing a basis for driving behavior decisions.

[0077] It should be understood that a preliminary trajectory can refer to a rough path generated based on driving behavior decisions and the global path (i.e., the macroscopic route from point A to point B), satisfying basic geometric and constraint conditions. It may only consider the starting point, ending point, and a few key points, but has not yet undergone fine-tuning. The preliminary trajectory clarifies the approximate scope and boundaries of the trajectory planning.

[0078] It should be explained that preset smoothness indicators can refer to mathematical standards and thresholds pre-set in the system to quantify whether a trajectory is smooth. Preset smoothness indicators may include the maximum value of curvature, the maximum value of the rate of change of curvature, and the maximum value of lateral acceleration, etc.

[0079] It should be noted that the polynomial function in this embodiment can be a fifth-order polynomial function. A fifth-order polynomial function is used to fit the initial trajectory, achieving smooth optimization of the initial trajectory. The mathematical expression of the fifth-order polynomial function is as follows: ; In the formula, y represents the lateral displacement, and x represents the longitudinal displacement. This is achieved by optimizing the coefficients. to This ensures that the generated trajectory curvature is continuous and without abrupt changes, fundamentally mimicking the smooth steering habits of human drivers and avoiding the "steering" phenomenon.

[0080] This embodiment discloses a method for conducting risk assessment based on the stated behavioral intent, determining the vehicle's driving behavior decision based on the risk assessment results, generating a preliminary trajectory based on the driving behavior decision and global path information, and then using a polynomial function to smooth and optimize the preliminary trajectory to generate a local trajectory that meets a preset smoothness index. Because this embodiment assesses the behavioral intent, generates a preliminary trajectory based on the driving behavior decision and global path information, and finally smooths and optimizes the preliminary trajectory to generate a local trajectory, compared to existing technologies, this embodiment improves the safety and comfort of trajectory planning.

[0081] Furthermore, this embodiment of the invention also proposes a storage medium storing an intelligent driving steering control program, which, when executed by a processor, implements the steps of the intelligent driving steering control method described above.

[0082] Reference Figure 7 , Figure 7 This is a structural block diagram of the first embodiment of the intelligent driving steering control device of the present invention.

[0083] like Figure 7 As shown, the intelligent driving steering control device proposed in this embodiment of the invention includes: a data processing module 701, an intent prediction module 702, an instruction generation module 703, and a control handover module 704.

[0084] The data processing module 701 is used to acquire multi-source perception data of the vehicle and perform fusion processing on the multi-source perception data to generate environmental perception data.

[0085] The intent prediction module 702 is used to predict the behavioral intent of dynamic traffic participants based on the environmental perception data.

[0086] The instruction generation module 703 is used to plan a local trajectory based on the behavioral intent and global path information, and generate steering control instructions based on the local trajectory using an adaptive model predictive control algorithm.

[0087] The control handover module 704 is used to generate a target steering control command based on the steering control command and the driver input when the driver's intention to take over is detected, and to perform steering control on the vehicle based on the target steering control command.

[0088] The data processing module 701 is further configured to acquire multi-source perception data of the vehicle, perform spatiotemporal synchronization processing on the multi-source perception data to obtain processed multi-source perception data, and use a deep learning model and Kalman filter to perform fusion processing on the processed multi-source perception data to generate environmental perception data.

[0089] The instruction generation module 703 is further configured to determine the target prediction model corresponding to the current control cycle from the model set based on the real-time state parameters of the vehicle and the road surface adhesion coefficient, wherein the model set includes multiple vehicle dynamics models; based on the target prediction model and the local trajectory, perform rolling optimization on the steering control sequence in the prediction time domain to obtain the optimal steering control sequence; and use the first control quantity in the optimal steering control sequence as the steering control instruction for the current control cycle.

[0090] The instruction generation module 703 is further configured to acquire the real-time state parameters and road surface adhesion coefficient of the vehicle; determine the adaptation weight of each vehicle dynamics model in the model set based on the real-time state parameters and the road surface adhesion coefficient; and fuse multiple vehicle dynamics models based on the adaptation weight to obtain the target prediction model corresponding to the current control cycle.

[0091] The control handover module 704 is also used to monitor the driver's visual characteristics, head posture, and hand torque in real time; based on the visual characteristics, head posture, and hand torque, determine whether the driver has an intention to take over; when the driver has an intention to take over, the steering control command and the driver input are weighted and fused based on a smooth handover strategy to generate a target steering control command; and the vehicle is steered according to the target steering control command.

[0092] This device embodiment discloses acquiring multi-source perception data of a vehicle and fusing the multi-source perception data to generate environmental perception data; predicting the behavioral intentions of dynamic traffic participants based on the environmental perception data; planning a local trajectory based on the behavioral intentions and global path information; and generating steering control commands based on the local trajectory using an adaptive model predictive control algorithm; when a driver's intention to take over is detected, generating a target steering control command based on the steering control command and driver input, and then performing steering control on the vehicle based on the target steering control command. Because this device embodiment fuses multi-source perception data, predicts the behavioral intentions of dynamic traffic participants, plans a local trajectory based on the behavioral intentions, generates steering control commands using an adaptive model predictive control algorithm, and finally generates a target steering control command based on the steering control command and driver input when a driver's intention to take over is detected, compared to existing technologies, this device embodiment improves the safety, comfort, and reliability of intelligent driving.

[0093] Based on the first embodiment of the intelligent driving steering control device of the present invention described above, a second embodiment of the intelligent driving steering control device of the present invention is proposed.

[0094] In this embodiment, the intent prediction module 702 is further configured to construct a traffic scene map based on the environmental perception data, wherein the nodes in the traffic scene map represent dynamic traffic participants, and the edges in the traffic scene map represent the spatial and interactive relationships between the dynamic traffic participants; input the traffic scene map into a preset graph neural network model to obtain the predicted trajectory and behavioral intent probability of each dynamic traffic participant; and use the predicted trajectory and behavioral intent probability of each dynamic traffic participant as the behavioral intent of each dynamic traffic participant.

[0095] Other embodiments or specific implementations of the intelligent driving steering control device of the present invention can be referred to the above-described method embodiments, and will not be repeated here.

[0096] This application provides an intelligent driving steering control device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the intelligent driving steering control method in the first embodiment described above.

[0097] The following is for reference. Figure 8 The diagram illustrates a structural schematic of an intelligent driving steering control device suitable for implementing embodiments of this application. The intelligent driving steering control device in this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 8 The intelligent driving steering control device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0098] like Figure 8As shown, the intelligent driving steering control device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the intelligent driving steering control device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. The communication device 1009 allows the intelligent driving steering control device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows an intelligent driving steering control device with various systems, it should be understood that it is not required to implement or possess all of the systems shown. More or fewer systems may be implemented alternatively.

[0099] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0100] The intelligent driving steering control device provided in this application, employing the intelligent driving steering control method described in the above embodiments, can solve the technical problems of insufficient safety, comfort, and reliability caused by the limited perception capabilities, insufficient predictive capabilities, and insufficient human-like control of existing intelligent driving steering control technologies in complex dynamic traffic environments. Compared with the prior art, the beneficial effects of the intelligent driving steering control device provided in this application are the same as those of the intelligent driving steering control method provided in the above embodiments, and other technical features of this intelligent driving steering control device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0101] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0102] 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.

[0103] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0104] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0105] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0106] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the protection scope of the present invention.

Claims

1. An intelligent driving steering control method, characterized by, The method comprises: obtaining multi-source perception data of the ego vehicle, and fusing the multi-source perception data to generate environment perception data; predicting behavior intention of dynamic traffic participants based on the environment perception data; planning a local trajectory according to the behavior intention and global path information, and generating a steering control instruction based on the local trajectory through an adaptive model predictive control algorithm; when detecting a takeover intention of the driver, generating a target steering control instruction according to the steering control instruction and the driver input, and performing steering control on the ego vehicle according to the target steering control instruction.

2. The intelligent driving steering control method of claim 1, wherein, The step of predicting the behavior intention of the dynamic traffic participants based on the environment perception data comprises: constructing a traffic scene graph based on the environment perception data, wherein the nodes in the traffic scene graph represent the dynamic traffic participants, and the edges in the traffic scene graph represent the spatial relationship and interaction relationship between the dynamic traffic participants; inputting the traffic scene graph into a preset graph neural network model to obtain the predicted trajectory and behavior intention probability of each dynamic traffic participant; taking the predicted trajectory and behavior intention probability of each dynamic traffic participant as the behavior intention of each dynamic traffic participant.

3. The intelligent driving steering control method of claim 1, wherein, The step of planning a local trajectory according to the behavior intention and global path information comprises: performing risk assessment based on the behavior intention, and determining the driving behavior decision of the ego vehicle according to the risk assessment result; generating a preliminary trajectory based on the driving behavior decision and global path information; performing smoothing optimization processing on the preliminary trajectory by using a polynomial function to generate a local trajectory that satisfies a preset smoothing degree index.

4. The intelligent driving steering control method of claim 1, wherein, The step of generating a steering control instruction based on the local trajectory through an adaptive model predictive control algorithm comprises: determining a target prediction model corresponding to the current control period from a model set according to the real-time state parameters of the ego vehicle and the road adhesion coefficient, wherein the model set comprises a plurality of vehicle dynamics models; performing rolling optimization on a steering control sequence in a prediction time domain based on the target prediction model and the local trajectory to obtain an optimal steering control sequence; taking the first control amount in the optimal steering control sequence as the steering control instruction of the current control period.

5. The intelligent driving steering control method of claim 4, wherein, The step of determining a target prediction model corresponding to the current control period from a model set according to the real-time state parameters of the ego vehicle and the road adhesion coefficient comprises: obtaining the real-time state parameters of the ego vehicle and the road adhesion coefficient; determining the adaptation weight of each vehicle dynamics model in the model set according to the real-time state parameters and the road adhesion coefficient; fusing a plurality of vehicle dynamics models according to the adaptation weight to obtain a target prediction model corresponding to the current control period.

6. The intelligent driving steering control method of claim 1, wherein, The step of generating a target steering control instruction according to the steering control instruction and the driver input when detecting a takeover intention of the driver, and performing steering control on the ego vehicle according to the target steering control instruction comprises: real-time monitoring the visual features, head posture and hand torque of the driver; determine whether the driver has a takeover intention based on the visual features, the head pose, and the hand moment; when the driver has a takeover intention, weight and fuse the steering control instruction and the driver input based on a control right smooth handover strategy to generate a target steering control instruction; perform steering control on the ego vehicle according to the target steering control instruction.

7. The intelligent driving steering control method of any one of claims 1-6, wherein, The step of obtaining multi-source perception data of the ego vehicle and performing fusion processing on the multi-source perception data to generate environment perception data includes: obtaining multi-source perception data of the ego vehicle, performing spatiotemporal synchronization processing on the multi-source perception data to obtain processed multi-source perception data; performing fusion processing on the processed multi-source perception data using a deep learning model and Kalman filtering to generate environment perception data.

8. An intelligent driving steering control device, characterized by comprising: The device includes: a data processing module configured to obtain multi-source perception data of the ego vehicle and perform fusion processing on the multi-source perception data to generate environment perception data; an intention prediction module configured to predict the behavior intention of a dynamic traffic participant based on the environment perception data; an instruction generation module configured to plan a local trajectory according to the behavior intention and global path information, and generate a steering control instruction based on the local trajectory using an adaptive model predictive control algorithm; a control handover module configured to, when a driver's takeover intention is detected, generate a target steering control instruction according to the steering control instruction and the driver input, and perform steering control on the ego vehicle according to the target steering control instruction.

9. An intelligent driving steering control apparatus characterized by comprising: The device includes a memory, a processor, and an intelligent driving steering control program stored on the memory and executable on the processor, the intelligent driving steering control program being configured to implement the steps of the intelligent driving steering control method according to any one of claims 1 to 7.

10. A storage medium, characterized by The storage medium stores an intelligent driving steering control program, and the intelligent driving steering control program, when executed by a processor, implements the steps of the intelligent driving steering control method according to any one of claims 1 to 7.

Images