Active and passive safety fusion control method and device for vehicle

Abstract

The invention provides an active and passive safety fusion control method and device for a vehicle, and the method comprises the steps: carrying out the preprocessing of real-time data collected by different sensors in the vehicle, building a unified environment model representing the surrounding environment of the vehicle based on the real-time data and the preprocessed data, vehicle collision risk prediction is carried out based on the unified environment model, the real-time data and preset three-dimensional calibration information, and an active braking instruction set and a passive safety instruction set are generated through a reinforcement learning mode based on a vehicle collision risk prediction result; an active safety weight and a passive safety weight are determined based on the real-time data and a vehicle collision risk prediction result; and based on the active safety weight and the passive safety weight, a vehicle active safety system is dispatched to execute a first target instruction in the active braking instruction set, and a vehicle passive safety system is dispatched to execute a second target instruction in the passive safety instruction set. According to the invention, the problem of insufficient cooperation degree of active and passive safety systems in the existing vehicle safety technology can be relieved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle safety technology, and in particular to a method and device for integrating active and passive safety control of a vehicle. Background Art

[0002] Among existing vehicle safety technologies, active safety technologies (such as anti-collision warning and automatic braking systems) and passive safety technologies (such as airbags and energy-absorbing structures) are usually independent of each other, and have the following main defects: insufficient information exchange between active and passive safety systems, and uncoordinated response timing; discretization of calibration parameters, and suboptimal overall safety performance; active and passive safety systems can only achieve coordination by cascading simple signals, but the deep coupling method is immature, especially difficult to meet vehicle safety requirements in multiple scenarios of autonomous driving. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a vehicle active and passive safety fusion control method and device to achieve coordinated control of the active and passive safety systems, improve the overall safety performance of the vehicle, and thus alleviate the problem of insufficient coordination between the active and passive safety systems in existing vehicle safety technologies.

[0004] In a first aspect, an embodiment of the present invention provides a method for fusion control of active and passive safety of a vehicle, comprising: preprocessing real-time data collected by different sensors in the vehicle; constructing a unified environmental model representing the vehicle's surrounding environment based on the real-time data and the preprocessed data, and predicting vehicle collision risk based on the unified environmental model, the real-time data and preset three-dimensional calibration information; generating an active braking instruction set and a passive safety instruction set through reinforcement learning based on the vehicle collision risk prediction results, and determining active safety weights and passive safety weights based on the real-time data and the vehicle collision risk prediction results; and scheduling the vehicle's active safety system to execute the first target instruction in the active braking instruction set and scheduling the vehicle's passive safety system to execute the second target instruction in the passive safety instruction set based on the active safety weight and the passive safety weight.

[0005] In the second aspect, an embodiment of the present invention also provides an active and passive safety fusion control device for a vehicle, comprising: a preprocessing module for preprocessing real-time data collected by different sensors in the vehicle; a modeling and prediction module for constructing a unified environmental model representing the vehicle's surrounding environment based on the real-time data and the preprocessed data, and predicting the vehicle collision risk based on the unified environmental model, the real-time data and preset three-dimensional calibration information; an instruction weight module for generating an active braking instruction set and a passive safety instruction set through reinforcement learning based on the vehicle collision risk prediction results, and determining the active safety weight and the passive safety weight based on the real-time data and the vehicle collision risk prediction results; an instruction execution module for scheduling the vehicle's active safety system to execute the first target instruction in the active braking instruction set and scheduling the vehicle's passive safety system to execute the second target instruction in the passive safety instruction set based on the active safety weight and the passive safety weight.

[0006] An embodiment of the present invention provides a vehicle active and passive safety fusion control method and device. First, real-time data collected by different sensors in the vehicle is preprocessed. Then, a unified environmental model representing the vehicle's surrounding environment is constructed based on the real-time data and the preprocessed data. The vehicle collision risk is predicted based on the unified environmental model, the real-time data, and preset three-dimensional calibration information. Then, an active braking instruction set and a passive safety instruction set are generated through reinforcement learning based on the vehicle collision risk prediction results. The active safety weight and the passive safety weight are determined based on the real-time data and the vehicle collision risk prediction results. Finally, based on the active safety weight and the passive safety weight, the vehicle active safety system is scheduled to execute the first target instruction in the active braking instruction set and the vehicle passive safety system is scheduled to execute the second target instruction in the passive safety instruction set. Using the above technology, the real-time data collected by different sensors in the vehicle and the preset three-dimensional calibration information can be used to adaptively schedule the vehicle's active safety system and the vehicle's passive safety system, which can achieve coordinated control of the vehicle's active and passive safety systems, thereby improving the overall safety performance of the vehicle.

[0007] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or understood by practicing the present invention. The objectives and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0008] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0010] Figure 1 Schematic diagram of a flow chart of a vehicle active and passive safety fusion control method according to an embodiment of the present invention; Figure 2 This is a diagram showing the overall architecture of the active and passive safety fusion control system of a vehicle in an embodiment of the present invention; Figure 3 A schematic diagram of a dynamic weight allocation algorithm according to an embodiment of the present invention; Figure 4 Schematic diagram of a three-dimensional calibration matrix data structure in an embodiment of the present invention; Figure 5 The figure is a schematic structural diagram of an active and passive safety fusion control device for a vehicle according to an embodiment of the present invention. DETAILED DESCRIPTION

[0011] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0012] At present, among the existing vehicle safety technologies, active safety technologies and passive safety technologies have the following main defects: insufficient information exchange between the active and passive safety systems, and uncoordinated response timing; discretization of calibration parameters, and suboptimal overall safety performance; active and passive safety systems can only achieve coordination by cascading simple signals, but the deep coupling method is immature, especially it is difficult to meet the vehicle safety needs in multiple scenarios of autonomous driving.

[0013] Based on this, the present invention provides a vehicle active and passive safety fusion control method and device, which can realize the coordinated control of the active and passive safety systems, improve the overall safety performance of the vehicle, and thus alleviate the problem of insufficient coordination between the active and passive safety systems in existing vehicle safety technologies.

[0014] To facilitate understanding of this embodiment, a vehicle active and passive safety fusion control method disclosed in an embodiment of the present invention is first introduced in detail. This method can be applied to a vehicle controller, especially a car controller, see Figure 1As shown, the method may include the following steps: Step S102: pre-processing the real-time data collected by different sensors in the vehicle.

[0015] Different sensors such as millimeter-wave radars, visual cameras, lidars, seat pressure sensors, acceleration sensors, speed sensors, angle sensors, position sensors, and meteorological sensors can be installed in vehicles to collect relevant data as real-time data. The real-time data collected by different sensors can then be pre-processed by filtering, clustering, interpolation, deduplication, time synchronization, and coordinate system conversion.

[0016] Step S104 : constructing a unified environment model representing the vehicle's surrounding environment based on the real-time data and the pre-processed data, and performing vehicle collision risk prediction based on the unified environment model, the real-time data, and the preset three-dimensional calibration information.

[0017] Step S106 , generating an active braking instruction set and a passive safety instruction set through reinforcement learning based on the vehicle collision risk prediction result, and determining an active safety weight and a passive safety weight based on real-time data and the vehicle collision risk prediction result.

[0018] Step S108 : Based on the active safety weight and the passive safety weight, respectively scheduling the vehicle active safety system to execute the first target instruction in the active braking instruction set and scheduling the vehicle passive safety system to execute the second target instruction in the passive safety instruction set.

[0019] An embodiment of the present invention provides a method for integrated control of active and passive safety of a vehicle. The method first pre-processes real-time data collected by different sensors in the vehicle, then constructs a unified environmental model representing the vehicle's surrounding environment based on the real-time data and the pre-processed data, and then predicts the vehicle's collision risk based on the unified environmental model, the real-time data, and preset three-dimensional calibration information. Furthermore, an active braking instruction set and a passive safety instruction set are generated through reinforcement learning based on the vehicle collision risk prediction results. Active safety weights and passive safety weights are determined based on the real-time data and the vehicle collision risk prediction results. Finally, based on the active safety weights and passive safety weights, the vehicle's active safety system is scheduled to execute the first target instruction in the active braking instruction set and the vehicle's passive safety system is scheduled to execute the second target instruction in the passive safety instruction set. Using the above technology, the real-time data collected by different sensors in the vehicle and the preset three-dimensional calibration information can be used to adaptively schedule the vehicle's active safety system and the vehicle's passive safety system, enabling coordinated control of the vehicle's active and passive safety systems, thereby improving the vehicle's overall safety performance.

[0020] As a possible implementation, real-time data may include real-time environmental data, real-time seat pressure data, real-time vehicle motion data, real-time vehicle driver operation data, etc. Based on this, the above-mentioned step S102 (i.e., preprocessing the real-time data collected by different sensors in the vehicle) may include: performing spatiotemporal alignment on the real-time data, and fusing the spatiotemporally aligned data; wherein the spatiotemporally aligned data may be time-synchronized data in the same vehicle coordinate system.

[0021] The purpose of this spatiotemporal alignment of real-time data is to eliminate differences in temporal and spatial coordinate systems between sensor data, providing a consistent foundation for subsequent data fusion. Spatial alignment can be accomplished by first performing intrinsic and extrinsic calibration of the sensor. Secondly, the coordinates of each sensor-detected target (such as a point cloud, bounding box, or feature point) are uniformly converted to the vehicle coordinate system based on the sensor's extrinsic matrix (including rotation and translation matrices). Finally, the precise motion trajectory of each target is back-projected into the vehicle coordinate system to eliminate point cloud distortion caused by vehicle motion and complete motion compensation. The time alignment operation method can be: first, use GPS / PPS signals or dedicated time synchronization chips to ensure hardware time synchronization, and at the same time, add software timestamps to the data points collected by each sensor; when the software timestamps of data points collected by different sensors at the same time are inconsistent, it is necessary to interpolate the vehicle posture at the corresponding time and use the vehicle motion model or the sensor itself to predict the data points at the corresponding time, and add the timestamp corresponding to the corresponding time to the predicted new data points, so as to predict (extrapolate) the original data points to the newly generated unified timestamp; finally, set a reasonable buffer for each sensor or each target, and wait for the original data points collected by the associated sensors to arrive at the corresponding buffer within the specified time window before performing alignment processing (that is, the above-mentioned operation of predicting the original data points to the newly generated unified timestamp, including vehicle posture interpolation and data point prediction).

[0022] In actual application, the above-mentioned operation method for fusing the spatiotemporally aligned data can be: first, performing data-level fusion on the spatiotemporally aligned data through point cloud and pixel fusion, pixel-radar fusion, and deep learning feature fusion; second, associating the data corresponding to the same target detected by different sensors to achieve target-level fusion, and fusing the status information (such as position, velocity, acceleration, category, etc.) of each target after the successful association of data, and completing track management (such as track initialization, track confidence scoring, track deletion, track prediction, etc.). Finally, through the implementation of complementary, redundancy, and collaborative strategies, more accurate, complete, and robust target detection results, target tracking results, and scene understanding results are obtained than those of a single sensor.

[0023] As a possible implementation, the real-time data may further include real-time meteorological data, etc., and the unified environment model may include the classification and attributes of target objects in the vehicle's surrounding environment, as well as static environment information and scene context information of the vehicle's surrounding environment. Based on this, the construction of the unified environment model representing the vehicle's surrounding environment based on the real-time data and the pre-processed data in step S104 may include: Step A1: perform target detection on the preprocessed data to obtain the classification and attributes of the target object.

[0024] Step A2: extract traffic elements from the pre-processed data to obtain static environment information.

[0025] Step A3: Extract scene context content from the preprocessed data based on the classification and attributes of the target object, static environmental information, and real-time meteorological data to obtain scene context information; wherein the scene context content may include: road type, weather conditions, traffic flow status, etc.

[0026] Step A4: Perform three-dimensional environment modeling based on the classification and attributes of the target object, the static environment information, and the scene context information to obtain a unified environment model.

[0027] In actual application, the operation method of constructing a unified environmental model may include: first, performing target detection, traffic element extraction and other operations on the fused data obtained by first aligning the real-time data in time and space and then fusing it, and extracting the scene context content in combination with real-time meteorological data to obtain a dynamic target list (including all perceived vehicles, pedestrians, cyclists, animals and other moving or potentially moving objects and the geometry, motion and other attributes of each object), a static environment (including information on the driving area and traffic elements such as lane lines, stop lines, and traffic lights) and a scene context (including road type, weather conditions, and traffic flow status), and then using the dynamic target list, static environment and scene context to perform three-dimensional environmental modeling, generate a structured and semantic three-dimensional environmental model as a unified environmental model, and update the data of the three-dimensional environmental model in real time.

[0028] As a possible implementation, the above-mentioned preset three-dimensional calibration information may include pre-input collision time series, collision intensity level and occupant posture parameters, and the vehicle collision risk prediction result may include collision risk level, expected collision time, collision type, potential collision object and predicted collision area and / or collision point; based on this, the vehicle collision risk prediction based on the unified environment model, real-time data and preset three-dimensional calibration information in the above-mentioned step S104 may include: determining the key input conditions of the unified environment model, real-time vehicle motion data and real-time vehicle driver operation data, and using the preset prediction model to use the key input conditions and preset three-dimensional calibration information to predict the vehicle collision risk; wherein, the preset prediction model may adopt one of the following: kinematic extrapolation model, probability model, interactive perception prediction model.

[0029] In actual application, the operation methods of vehicle collision risk prediction may include: setting key input conditions such as a unified environmental model, precise vehicle status (such as vehicle acceleration, vehicle speed, vehicle angular velocity, etc. detected by sensors), and driver operation (such as steering wheel rotation, braking, accelerator, etc. detected by sensors), selecting one of the kinematic extrapolation model, probability model, interactive perception prediction model and other models as the vehicle collision risk prediction model, and when the set key input conditions are met, inputting the collision time series, collision intensity level, occupant posture parameters, etc. into the vehicle collision risk prediction model for prediction. The vehicle collision risk prediction model outputs the collision risk level, expected collision time, collision type, potential collision objects, and collision area and / or collision point.

[0030] As a possible implementation, generating an active braking instruction set and a passive safety instruction set through reinforcement learning based on the vehicle collision risk prediction results in the above-mentioned step S106 may include: using a preset reinforcement learning model to output active braking instructions and passive safety instructions using the vehicle collision risk prediction results; forming the obtained active braking instructions into an active braking instruction set, and forming the obtained passive safety instructions into a passive safety instruction set.

[0031] In actual application, the operation method of generating active braking instruction sets and passive safety instruction sets may include: pre-training a reinforcement learning model and deploying and configuring the reinforcement learning model on a reinforcement learning decision engine; after obtaining the vehicle collision risk prediction results (including collision risk level, expected collision time, collision type, potential collision objects, collision area and / or collision point, etc.), inputting the vehicle collision risk prediction results into the reinforcement learning decision engine for calculation. The reinforcement learning decision engine simultaneously outputs active braking (such as AEB, etc.) instructions and passive safety (such as airbags, etc.) instructions to avoid the problem of conflicting responses between active and passive safety systems in traditional vehicle safety technology. The obtained active braking instructions constitute the active braking instruction set, and the obtained passive safety instructions constitute the passive safety instruction set.

[0032] As a possible implementation, the real-time environmental data may include real-time radar data, real-time image data, etc., and the real-time vehicle motion data may include real-time vehicle acceleration data, real-time vehicle speed data, real-time vehicle angular velocity data, etc. Based on this, the active safety weight and the passive safety weight determined based on the real-time data and the vehicle collision risk prediction result include: Step a1: determining the real-time occupant status based on the real-time image data and / or the real-time seat pressure data, and determining the operating parameters of the road surface on which the vehicle is located based on the real-time vehicle motion data.

[0033] Among them, the operating parameters may include road adhesion coefficient, slope, etc. The road adhesion coefficient can be obtained by reverse deduction using a preset dynamic model using real-time vehicle motion data, and the slope can be determined based on real-time vehicle acceleration data and real-time vehicle speed data.

[0034] In actual application, the type and posture of the occupant can be identified through seat pressure detected by the seat pressure sensor and / or images of the occupant detected by the visual camera. The road adhesion coefficient can be obtained by real-time reverse engineering based on the tire-road dynamics model. Specifically, the system uses sensors to acquire signals such as wheel speed, longitudinal acceleration, lateral acceleration, yaw rate, steering angle, master cylinder pressure, and engine torque in real time. A vehicle state observer is then used to estimate key vehicle states (such as longitudinal speed and sideslip angle) in real time. Tire forces are calculated based on dynamic equations (such as wheel speed dynamics and vehicle dynamics). The dynamic vertical load on each tire is estimated based on the acceleration signal and suspension model. A determination is made as to whether a valid dynamic excitation is present. If so, the road adhesion coefficient is then reverse engineered based on the obtained tire forces and dynamic vertical loads. The slope can be obtained through the fusion of acceleration and velocity differentials. The specific operation method is as follows: the vehicle's acceleration and velocity are obtained in real time through sensors, a state space model (including state variables, state equations, and observation equations) is constructed, and based on the state space model, the acceleration and velocity are fused and calculated using the differential of the velocity as a reference benchmark through a Kalman filter to obtain the slope.

[0035] Step a2: determining the active safety weight based on the real-time vehicle motion data, the road adhesion coefficient, the real-time occupant status and the vehicle collision risk prediction result.

[0036] Exemplarily, the above-mentioned vehicle collision risk prediction results may also include the collision probability of the collision type, the above-mentioned real-time occupant status may include the real-time occupant type and the real-time occupant posture, the adhesion coefficient influencing factor may be determined based on the road adhesion coefficient, and the risk sensitive factor may be determined based on the real-time vehicle motion data, the collision probability and the real-time occupant status; the active safety weight may be determined based on the risk sensitive factor, the collision probability and the adhesion coefficient influencing factor.

[0037] In actual application, the road adhesion coefficient can be Determine the factors affecting the adhesion coefficient ; According to the collision energy E (reflected by the vehicle speed), the collision probability of the collision type and occupant type data are used to obtain the risk sensitivity factor (which can be recorded as ), for example, when the collision energy E>50kJ (equivalent to a 30km / h collision) Based on risk-sensitive factors , collision probability and adhesion coefficient influencing factors The active safety weight is calculated using the following formula: :

[0038] in, is the risk offset, The value range is .

[0039] Step a3: Determine the passive safety weight based on the real-time vehicle motion data, road adhesion coefficient, slope, and vehicle collision risk prediction results.

[0040] For example, a slope compensation may be determined based on the slope, and an urgency factor may be determined based on the estimated collision time, the road adhesion coefficient, and real-time vehicle motion data; and a passive safety weight may be determined based on the urgency factor, the estimated collision time, and the slope compensation.

[0041] In actual application, the slope Determining slope compensation ; According to the expected collision time TTC, road adhesion coefficient The urgency factor (which can be expressed as ), The value range is ; Based on the urgency factor , Estimated Time to Collision TTC and Slope Compensation The passive safety weight is calculated using the following formula: :

[0042] in, The value range is .

[0043] As a possible implementation, the above-mentioned step S108 (i.e., scheduling the vehicle's active safety system to execute the first target instruction in the active braking instruction set and scheduling the vehicle's passive safety system to execute the second target instruction in the passive safety instruction set based on the active safety weight and the passive safety weight) may include: using a preset timing arbiter to use the active safety weight and the passive safety weight to determine the respective scheduling timings of the first target instruction and the second target instruction, and scheduling the first actuator corresponding to the vehicle's active safety system to execute the first target instruction and scheduling the second actuator corresponding to the vehicle's passive safety system to execute the second target instruction according to the scheduling timings.

[0044] In actual application, priorities can also be set for the instructions that need to be dispatched to execute the vehicle's active and passive safety systems. Common instruction priority setting strategies are shown in Table 1. Bandwidth allocation can be achieved through the CAN FD bus based on instruction priority, combined with high-speed data field transmission (up to 5Mbps) and dynamic load management mechanisms, to send the instructions that need to be dispatched to execute the vehicle's active and passive safety systems to the actuators of the vehicle's active and passive safety systems for execution.

[0045] Table 1 Common instruction priority setting strategies

[0046] In practical applications, to further ensure vehicle safety, a vehicle system operating mode switching mechanism can be introduced, specifically dividing the vehicle system into multiple different operating modes. Table 2 shows the classification of vehicle system operating modes: local full perception mode, local degraded mode, V2X collaborative perception mode, and minimum risk state mode. Table 2 also shows the trigger conditions, perception dependencies, and control permissions for each operating mode. For example, when a sensor fails, it automatically switches to V2X collaborative perception mode. This is defined based on dynamic calibration of functional safety levels, ISO21434 cybersecurity constraints, and a quantitative perception capability model to enhance functional safety, collaboratively defend against cybersecurity, and optimize system efficiency. Table 2 shows the classification of system operating modes.

[0047] Table 2 Classification of vehicle system working modes

[0048] For ease of understanding, the implementation of the active and passive safety fusion control method of the above-mentioned vehicle is described below by taking a specific application as an example.

[0049] A vehicle active and passive safety fusion control system can be constructed to implement the above vehicle active and passive safety fusion control method. The overall architecture of the vehicle active and passive safety fusion control system (data interaction path of active and passive safety systems) is as follows: Figure 2 As shown in Figure 2, this architecture uses a layered design to efficiently transform multimodal data into executable instructions. It features multi-sensor spatiotemporal synchronization (error <1ms), 360° environmental perception output (update rate 100Hz), real-time decision-making based on reinforcement learning (latency <10ms), dynamic weights α / β satisfying the α+β=1 constraint, nanosecond-level instruction synchronization (e.g., airbag and seatbelt pretensioning timing deviation <0.5ms), and support for ASIL D functional safety downgrade.

[0050] See also Figure 2As shown in the figure, the overall architecture of the vehicle's active and passive safety fusion control system includes a sensor fusion layer, a decision-making collaboration layer, and an execution coordination layer. The architecture includes different sensors in the vehicle (such as millimeter-wave radar, visual camera, lidar, seat pressure sensor, acceleration sensor, speed sensor, angle sensor, position sensor, meteorological sensor, etc.), a spatiotemporal alignment module, a multi-source data fusion engine, a unified environment model generation module, a collision risk prediction module, a reinforcement learning decision engine, a dynamic weight allocator, a timing arbitrator, a hardware driver interface, and a CAN FD bus. Each sensor is connected to the spatiotemporal alignment module, and the spatiotemporal alignment module, the multi-source data fusion engine, the unified environment model generation module, the collision risk prediction module, the reinforcement learning decision engine, the dynamic weight allocator, the timing arbitrator, the actuator coordinator, and the hardware driver interface are connected in sequence. The hardware driver interface is connected to the actuators of the active and passive safety systems through the CAN FD bus.

[0051] See also Figure 2 As shown in the figure, the working principle of the vehicle's active and passive safety fusion control system is as follows: The spatiotemporal alignment module performs spatiotemporal alignment on the real-time data collected by different sensors in the vehicle, eliminates the differences in the time and space coordinate systems of different sensor data, and provides a consistency basis for subsequent data fusion; the multi-source data fusion engine fuses the spatiotemporal aligned data (including data-level fusion, target-level fusion, fusion target status information, track management, etc.), and outputs target detection results, target tracking results, and scene understanding results that are more accurate, complete, and robust than those of a single sensor; the unified environment model generation module integrates the results output by the multi-source data fusion engine (such as target detection, traffic element extraction, scene context content extraction, etc.), and uses the integrated results (including dynamic target list, static environment and scene context) to perform three-dimensional environment modeling to generate a structured and semantic unified environment model The collision risk prediction module uses a pre-defined three-dimensional calibration matrix (including collision time series, collision intensity level, and occupant posture parameters) to predict vehicle collision risk using one of the kinematic extrapolation model, probability model, and interactive perception prediction model, provided that key input conditions such as a unified environmental model, precise vehicle status, and driver operation are met. The module then outputs the vehicle collision risk prediction results (including collision risk level, expected collision time, collision type and collision probability, potential collision objects, and collision area and / or collision point). The reinforcement learning decision engine uses the vehicle collision risk prediction results for calculations and simultaneously outputs active braking commands and passive safety commands to avoid conflicting responses between the vehicle's active and passive safety systems, thereby obtaining an active braking command set and a passive safety command set. The dynamic weight allocator (including the weight calculation engine), timing arbitrator, actuator coordinator, hardware driver interface and CAN FD bus as well as the actuators of active and passive safety systems work together to implement the dynamic weight allocation algorithm. The core logic of the dynamic weight allocation algorithm is as follows: Figure 3 shown; see Figure 3 As shown in Figure 2, the workflow of the dynamic weight allocation algorithm can be summarized into the following key steps: (1.1) The weight calculation engine obtains collision probability prediction results, occupant status monitoring results, and environmental perception results; Collision probability prediction: This system processes data collected by sensors such as millimeter-wave radar, visual cameras, lidar, seat pressure sensors, accelerometers, speed sensors, angle sensors, position sensors, and meteorological sensors (including raw point cloud filtering, target clustering, motion state estimation, TTC calculation, probability mapping, and collision risk prediction based on the vehicle coordinate system) to ensure spatiotemporal alignment and coordinate unification, thereby resolving conflicting data. Occupant status monitoring: Identifying the occupant type and posture as occupant status through seat pressure and / or images captured by visual cameras; Environmental perception: Acquires operating parameters such as road adhesion coefficient and slope. The road adhesion coefficient is derived through real-time inverse calculation based on the tire-road dynamics model, and the slope is derived through acceleration-velocity differential fusion. (1.2) The weight calculation engine uses nonlinear functions to dynamically generate active safety weights and passive safety weights. The classification and weights of nonlinear functions are shown in Table 3. Table 3 Nonlinear function classification and weights

[0052] (1.3) The timing arbiter uses the active safety weight and the passive safety weight to schedule the timing of the target instructions to be executed in the active braking instruction set and the passive safety instruction set. The actuator coordinator schedules the actuators of the vehicle's active and passive safety systems to execute the target instructions based on the timing scheduling results through the hardware driver interface and the CAN FD bus. Among them, the timing arbiter serves as a resource scheduler in the time dimension to resolve the execution timing conflicts of multiple instructions. It is usually implemented as ASIL-D level in the ISO26262 architecture. The actuator coordinator serves as an action optimization in the spatial dimension to coordinate the physical action compatibility of multiple actuators. It is usually implemented as ASIL-B level in the ISO 26262 architecture. The classification of typical weight distribution scenarios is shown in Table 4.

[0053] Table 4 Classification of typical weight distribution scenarios

[0054] The above three-dimensional calibration matrix data structure is as follows Figure 4 See Figure 4 As shown in the figure, the three-dimensional calibration matrix data consists of three dimensions: longitudinal, lateral and depth. The information corresponding to the longitudinal dimension is the TTC time series (including the key time nodes before the collision and the time windows divided by these time nodes), the information corresponding to the lateral dimension is the collision intensity level (obtained according to the vehicle speed), and the information corresponding to the depth dimension includes occupant posture parameters (including weight percentile, sitting angle, seat belt status, etc.).

[0055] See also Figure 4 As shown in Figure 5, the information contained in the three-dimensional calibration matrix covers the critical time window before collision (as shown in Table 5), realizing the calibration of the entire process from warning (5.0s) to inevitable collision (0.0s).

[0056] Table 5 Critical time window before collision

[0057] See also Figure 4 As shown in Table 6, based on the three-dimensional calibration matrix, the collision intensity level classification can be carried out according to different NCAP standard working conditions to be compatible with different collision energy scenarios.

[0058] Table 6 Collision strength level classification

[0059] See also Figure 4 As shown in Figure 7, based on the three-dimensional calibration matrix, key occupant protection parameters can be optimized for occupant diversity, providing balanced protection for occupants such as the 5th percentile female (small female), the 50th percentile male (medium male), and the 95th percentile male (large male). This is achieved primarily by adjusting the following three key protection parameters: airbag deployment timing (the time from trigger command to full deployment, typically 15-50ms, which determines the optimal cushioning state when the occupant contacts the airbag. If the airbag deploys too early, the airbag will have already deflated by the time the occupant contacts the airbag; if it deploys too late, the occupant will have already struck the steering wheel before contact); seatbelt force limiter (controls the release speed of the webbing through a predefined mechanical threshold to prevent excessive chest compression); and pretensioner trigger strength (the amount of pretensioner force applied when the pretensioner is activated, instantly retracting the seatbelt to eliminate webbing slack and reduce forward motion). The strategy for adjusting key occupant protection parameters for weight percentiles is shown in Table 7.

[0060] Table 7 Occupant protection key parameter adjustment strategy

[0061] See also Figure 4As shown in the figure, based on the three-dimensional calibration matrix, the resolution of the critical pre-crash time window can be dynamically adjusted: the time window with a TTC less than 1.0s is increased to 0.1s, and the time window with a TTC greater than 1.0s is increased to 0.5s. Based on the three-dimensional calibration matrix, the calibration process can be directly linked to regulations and NCAP assessments to ensure compliance. Based on the three-dimensional calibration matrix, out-of-position occupant protection deficiencies can be compensated for by adjusting the seating angle (-15° to +25°). The out-of-position occupant risk and compensation logic are shown in Table 8.

[0062] Table 8 Risks and compensation logic for out-of-position occupants

[0063] See also Figures 1 to 4 As shown, combined with the above content, taking a certain car model as an example, some implementation methods of the active and passive safety fusion control method of the above vehicle are exemplarily described as follows.

[0064] During the pre-crash phase (TTC = 2s): millimeter-wave radar data and visual data (i.e., image data collected by the visual camera) are temporally aligned and fused to generate a unified environmental model. This model is then combined with a three-dimensional calibration matrix to predict collision risk. A dynamic weight allocation algorithm is then calculated to coordinate the execution of instructions by the active and passive safety systems. When a collision risk is predicted, the active safety system triggers brake preload. Seatbelt preload parameters are adjusted based on different collision probabilities, occupant types, and triggering timing (as shown in Table 7). During the unavoidable collision phase (TTC = 0.3s): The occupant position is monitored through the seat pressure matrix (seat pressure information expressed in matrix form), the in-vehicle visual camera, and the seatbelt buckle sensor. The airbag deployment timing is controlled in multiple stages according to different collision scenarios and seatbelt airbag firing strategies (as shown in Table 7). The braking system is controlled to maintain a deceleration of 0.8g to reduce the collision intensity and energy.

[0065] The beneficial effects of the active and passive safety fusion control method of the above-mentioned vehicle can be mainly reflected in: establishing a joint control domain of active and passive safety, designing a three-layer interactive architecture of "sensor fusion layer + decision-making collaboration layer + execution coordination layer", introducing a dynamic weight distribution algorithm, and adjusting the participation of the active and passive safety systems in real time according to the collision probability, so as to realize the coordinated optimization of the vehicle's active and passive safety systems and improve the overall safety performance; defining a three-dimensional calibration matrix including TTC time series, collision intensity level and occupant posture parameters for vehicle collision risk prediction, thereby improving the comprehensiveness of vehicle collision risk prediction, thereby facilitating the reliability of the calculation results of the dynamic weight distribution algorithm, thereby facilitating the coordination of the vehicle's active and passive safety systems to reduce occupant injuries and enhance the coordination of the responses of the vehicle's active and passive safety systems.

[0066] Based on the above-mentioned active and passive safety fusion control method of a vehicle, an embodiment of the present invention further provides an active and passive safety fusion control device for a vehicle, see Figure 5 As shown, the device may include the following modules: The pre-processing module 502 is used to pre-process the real-time data collected by different sensors in the vehicle.

[0067] The modeling and prediction module 504 is used to build a unified environmental model that represents the vehicle's surrounding environment based on the real-time data and the pre-processed data, and to predict the vehicle collision risk based on the unified environmental model, the real-time data and the preset three-dimensional calibration information.

[0068] The instruction weight module 506 is used to generate an active braking instruction set and a passive safety instruction set through reinforcement learning based on the vehicle collision risk prediction result, and determine the active safety weight and the passive safety weight based on the real-time data and the vehicle collision risk prediction result.

[0069] The instruction execution module 508 is used to schedule the vehicle active safety system to execute the first target instruction in the active braking instruction set and schedule the vehicle passive safety system to execute the second target instruction in the passive safety instruction set based on the active safety weight and the passive safety weight.

[0070] An embodiment of the present invention provides a vehicle active and passive safety fusion control device that can utilize real-time data collected by different sensors in the vehicle and preset three-dimensional calibration information to adaptively schedule the vehicle's active safety system and vehicle passive safety system, thereby achieving coordinated control of the vehicle's active and passive safety systems, thereby improving the vehicle's overall safety performance.

[0071] The active and passive safety fusion control device for a vehicle provided in an embodiment of the present invention has the same implementation principle and technical effects as those in the aforementioned embodiment of the active and passive safety fusion control method for a vehicle. For the sake of brief description, for matters not mentioned in the embodiment of the active and passive safety fusion control device for a vehicle, reference may be made to the corresponding contents in the aforementioned embodiment of the active and passive safety fusion control method for a vehicle.

[0072] Unless otherwise specifically stated, the relative steps, numerical expressions and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present invention.

[0073] If the functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0074] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0075] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or replace some of the technical features therein with equivalents. Such modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A vehicle active and passive safety fusion control method, characterized in that: include: Pre-process the real-time data collected by different sensors in the vehicle; constructing a unified environmental model representing the vehicle's surrounding environment based on the real-time data and the pre-processed data, and performing a vehicle collision risk prediction based on the unified environmental model, the real-time data, and preset three-dimensional calibration information; generating an active braking instruction set and a passive safety instruction set based on the vehicle collision risk prediction result through reinforcement learning, and determining an active safety weight and a passive safety weight based on the real-time data and the vehicle collision risk prediction result; Based on the active safety weight and the passive safety weight, the vehicle active safety system is scheduled to execute the first target instruction in the active braking instruction set, and the vehicle passive safety system is scheduled to execute the second target instruction in the passive safety instruction set.

2. The active and passive safety fusion control method for a vehicle according to claim 1, characterized in that: The real-time data includes real-time environmental data, real-time seat pressure data, real-time vehicle motion data and real-time vehicle driver operation data; Preprocessing of real-time data collected by different sensors in the vehicle, including: The real-time data is time-space aligned and the time-space aligned data is fused; wherein the time-space aligned data is time-synchronized data in the same vehicle coordinate system.

3. The active and passive safety fusion control method for a vehicle according to claim 2, characterized in that: The preset three-dimensional calibration information includes a pre-input collision time sequence, collision intensity level, and occupant posture parameters; the vehicle collision risk prediction result includes a collision risk level, expected collision time, collision type, potential collision object, and predicted collision area and / or collision point; The vehicle collision risk prediction is performed based on the unified environment model, the real-time data, and the preset three-dimensional calibration information, including: Determine the key input conditions of the unified environment model, the real-time vehicle motion data, and the real-time vehicle driver operation data, and use a preset prediction model to predict the vehicle collision risk using the key input conditions and the preset three-dimensional calibration information; wherein the preset prediction model uses one of the following: a kinematic extrapolation model, a probability model, and an interactive perception prediction model.

4. The active and passive safety fusion control method for a vehicle according to claim 3, characterized in that: The real-time environmental data includes real-time radar data and real-time image data, and the real-time vehicle motion data includes real-time vehicle acceleration data, real-time vehicle speed data and real-time vehicle angular velocity data; Determining an active safety weight and a passive safety weight based on the real-time data and the vehicle collision risk prediction result includes: determining a real-time occupant status based on the real-time image data and / or the real-time seat pressure data, and determining operating condition parameters of a road surface on which the vehicle is located based on the real-time vehicle motion data; wherein the operating condition parameters include a road surface adhesion coefficient and a slope; the road surface adhesion coefficient is obtained by reverse engineering the real-time vehicle motion data using a preset dynamic model; and the slope is determined based on the real-time vehicle acceleration data and the real-time vehicle speed data; determining the active safety weight based on the real-time vehicle motion data, the road adhesion coefficient, the real-time occupant status, and the vehicle collision risk prediction result; The passive safety weight is determined based on the real-time vehicle motion data, the road adhesion coefficient, the slope, and the vehicle collision risk prediction result.

5. The active and passive safety fusion control method for a vehicle according to claim 4, characterized in that: The vehicle collision risk prediction result also includes the collision probability of the collision type, and the real-time occupant status includes the real-time occupant type and the real-time occupant posture; Determining the active safety weight based on the real-time vehicle motion data, the road adhesion coefficient, the real-time occupant status, and the vehicle collision risk prediction result includes: determining an adhesion coefficient influencing factor based on the road adhesion coefficient, and determining a risk sensitivity factor based on the real-time vehicle motion data, the collision probability, and the real-time occupant status; The active safety weight is determined based on the risk sensitivity factor, the collision probability, and the adhesion coefficient influencing factor.

6. The active and passive safety fusion control method for a vehicle according to claim 4, characterized in that: Determining the passive safety weight based on the real-time vehicle motion data, the road adhesion coefficient, the slope, and the vehicle collision risk prediction result includes: determining a slope compensation based on the slope and determining an urgency factor based on the estimated time to collision, the road surface adhesion coefficient, and the real-time vehicle motion data; The passive safety weight is determined based on the urgency factor, the estimated time to collision, and the slope compensation.

7. The active and passive safety fusion control method for a vehicle according to claim 1, characterized in that: Based on the vehicle collision risk prediction results, the active braking instruction set and passive safety instruction set are generated through reinforcement learning, including: Using a preset reinforcement learning model and the vehicle collision risk prediction result to output active braking instructions and passive safety instructions; The obtained active braking instructions are combined into the active braking instruction set, and the obtained passive safety instructions are combined into the passive safety instruction set.

8. The active and passive safety fusion control method for a vehicle according to claim 1, characterized in that: Based on the active safety weight and the passive safety weight, respectively scheduling the vehicle active safety system to execute the first target instruction in the active braking instruction set and scheduling the vehicle passive safety system to execute the second target instruction in the passive safety instruction set, including: A preset timing arbiter is used to determine the respective scheduling timings of the first target instruction and the second target instruction using the active safety weight and the passive safety weight, and the first actuator corresponding to the vehicle's active safety system is respectively scheduled to execute the first target instruction and the second actuator corresponding to the vehicle's passive safety system is respectively scheduled to execute the second target instruction according to the scheduling timings.

9. The active and passive safety fusion control method for a vehicle according to claim 2, characterized in that: The real-time data also includes real-time meteorological data; the unified environment model includes the classification and attributes of target objects in the vehicle's surrounding environment, as well as static environment information and scene context information of the vehicle's surrounding environment; A unified environment model representing the vehicle's surrounding environment is constructed based on the real-time data and the pre-processed data, including: Performing target detection on the preprocessed data to obtain the classification and attributes of the target object; Extracting traffic elements from the preprocessed data to obtain the static environment information; Extracting scene context from the preprocessed data based on the classification and attributes of the target object, the static environmental information, and the real-time meteorological data to obtain the scene context information; wherein the scene context includes at least one of the following: road type, weather conditions, and traffic flow status; Three-dimensional environment modeling is performed based on the classification and attributes of the target object, the static environment information, and the scene context information to obtain the unified environment model.

10. A vehicle active and passive safety fusion control device, characterized in that: include: A preprocessing module is used to preprocess the real-time data collected by different sensors in the vehicle; a modeling and prediction module, configured to construct a unified environmental model representing the vehicle's surrounding environment based on the real-time data and the preprocessed data, and to predict the vehicle collision risk based on the unified environmental model, the real-time data, and preset three-dimensional calibration information; An instruction weight module, configured to generate an active braking instruction set and a passive safety instruction set by reinforcement learning based on a vehicle collision risk prediction result, and determine an active safety weight and a passive safety weight based on the real-time data and the vehicle collision risk prediction result; The instruction execution module is used to schedule the vehicle active safety system to execute the first target instruction in the active braking instruction set and schedule the vehicle passive safety system to execute the second target instruction in the passive safety instruction set based on the active safety weight and the passive safety weight.

Images