Vehicle pre-judgment guiding method and device

Abstract

The invention discloses a vehicle pre-judgment guiding method and device. The method comprises the steps that vehicle state data, driver dominant intention data, environmental perception data, map context data and navigation path data are acquired; determining a hypothetical trajectory set of the vehicle based on the vehicle state data, the driver dominant intention data, the map context data and the navigation path data; for each dynamic target in the environmental perception data, determining a hypothetical trajectory set of the dynamic target; determining a plurality of potential conflict points based on the hypothetical trajectory set of the vehicle and the hypothetical trajectory set of each dynamic target; constructing a dynamic risk field based on the plurality of potential conflict points; determining a safety path based on the dynamic risk field; determining a guidance instruction based on the predicted collision time and severity of the plurality of potential conflict points; and rendering the safety path, the dynamic risk field and the guidance instruction into an A-column display screen of the vehicle. According to the method, the pre-judgment and decision-making assistance capability of the vehicle safety system is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of vehicle safety technology, and in particular to a vehicle prediction and guidance method and device. Background Technology

[0002] Currently, in the field of vehicle active safety, blind spot monitoring and warning is one of the key technologies for improving driving safety. The mainstream technical solutions mainly involve adding sensors (such as cameras and radar) to the vehicle to detect obstacles in blind spots and issuing warnings to the driver through visual or auditory signals.

[0003] However, these technological approaches generally suffer from a fundamental limitation: they operate on a "perception-response" model. That is, the system only alerts or warns the driver when it detects a risky target (such as a vehicle or pedestrian) entering or very close to the vehicle's blind spot. This model has several significant drawbacks in practice: first, the warning is delayed, requiring the target to enter the blind spot or the risk level to be reached before issuing an alert, leaving the driver with insufficient reaction time; second, the cognitive load is high, with most existing technologies directly displaying raw images or abstract icons, forcing the driver to interpret multiple target information, easily leading to information overload; third, the function is passive, merely acting as a hazard indicator, unable to proactively answer key questions such as whether to turn and how to safely pass based on predictions of the vehicle's intentions and the future behavior of surrounding targets, lacking proactive guidance capabilities.

[0004] Therefore, while existing blind spot safety technologies have made progress in terms of visualization, they have encountered bottlenecks in terms of intelligence and predictability, and there is an urgent need for a fundamental improvement solution that can realize the transformation from passive alarm to active prediction and from information display to decision support. Summary of the Invention

[0005] In order to improve the lag and deficiencies of existing early warning systems and enhance the prediction and decision-making support capabilities of vehicle safety systems, this invention provides a vehicle prediction guidance method and device.

[0006] In a first aspect, embodiments of the present invention provide a vehicle prediction and guidance method, which may include: Acquire vehicle status data, driver explicit intent data, environmental perception data, map context data, and navigation path data; Based on the vehicle status data, the driver's explicit intent data, the map context data, and the navigation path data, a set of hypothetical trajectories for the vehicle is determined; For each dynamic target in the environmental perception data, determine a set of hypothetical trajectories for the dynamic target; Based on the hypothetical trajectory set of the vehicle and the hypothetical trajectory set of each dynamic target, multiple potential conflict points are identified. A dynamic risk field is constructed based on the aforementioned multiple potential conflict points; Based on the dynamic risk field, a safe path is determined; Based on the predicted collision times and severity of the multiple potential conflict points, guidance instructions are determined. The safe path, the dynamic risk field, and the guidance instructions are rendered onto the vehicle's A-pillar display screen.

[0007] In one or more optional embodiments of this application, rendering the safe path, the dynamic risk field, and the guidance instructions onto the vehicle's A-pillar display screen includes: The safe path, the dynamic risk field, and the guidance instructions are converted into a guidance image; By using spatial registration and perspective fusion technology, the guide image is displayed on the vehicle's A-pillar display screen, so that the guide image is visually spatially aligned with the real external scene observed by the vehicle driver through the A-pillar.

[0008] In one or more optional embodiments of this application, converting the security path, the dynamic risk field, and the guidance instructions into a guidance image includes: The safe path is rendered as a first guide graphic; the first guide graphic is a semi-transparent filled strip area with a preset safety color, the transparency of which is determined according to the predicted collision time of potential conflict points in the safe path, and the edges are provided with dynamic flowing light effects to indicate the driving direction. The high-risk area in the dynamic risk field with a risk value higher than a preset risk threshold is rendered as a second guide graphic; the second guide graphic is a semi-transparent cloud-shaped or mesh-shaped area filled with a preset warning color, the transparency of which is determined according to the risk value of the high-risk area, and is accompanied by a periodic pulsating warning effect. The boot instructions are rendered as a third boot graphic; The guide image is obtained based on the first guide graphic, the second guide graphic, and the third guide graphic.

[0009] In one or more optional embodiments of this application, determining the guidance instruction based on the predicted collision time and severity of the plurality of potential conflict points includes: For each potential conflict point, if the predicted collision time of the potential conflict point is less than a first preset warning time or the severity is greater than a preset severity threshold, a strong intervention guidance instruction is generated; the strong intervention guidance instruction includes avoidance suggestions. If the predicted collision time of the potential conflict point is greater than or equal to the first preset warning time and less than the second preset warning time, a suggested guidance instruction is generated; the suggested guidance instruction includes a recommended safe driving area; the first preset warning time is less than the second preset warning time.

[0010] In one or more optional embodiments of this application, determining the set of hypothetical trajectories for the vehicle based on the vehicle state data, the driver's explicit intent data, the map context data, and the navigation path data includes: Based on the vehicle status data, a short-term trajectory is obtained; Based on the driver's explicit intent data, the map context data, and the navigation path data, driving intent distribution data is obtained; Based on the driving intention distribution data, a set of long-term hypothetical trajectories is determined; Based on the short-term trajectory and the long-term hypothetical trajectory set, the hypothetical trajectory set of the vehicle is obtained; the time length of each hypothetical trajectory in the hypothetical trajectory set of the vehicle is a preset future duration.

[0011] In one or more optional embodiments of this application, the driving intention distribution data includes multiple probability values ​​corresponding to driving intentions; The determination of the long-term hypothetical trajectory set based on the driving intention distribution data includes: Based on the probability value corresponding to each driving intention in the driving intention distribution data, multiple possible driving intentions are determined; Determine the long-term hypothetical trajectory corresponding to each possible driving intention; For each of the long-term hypothetical trajectories, the probability value corresponding to the long-term hypothetical trajectory is determined based on the probability value of the driving intention corresponding to the long-term hypothetical trajectory and the dynamic feasibility of the long-term hypothetical trajectory. By combining multiple long-term hypothesis trajectories and the probability value corresponding to each long-term hypothesis trajectory, the set of long-term hypothesis trajectories is obtained.

[0012] In one or more optional embodiments of this application, determining the set of hypothetical trajectories for each dynamic target in the environmental perception data includes: If the category of the dynamic target is a vehicle and there are other vehicles in front of the dynamic target, then the set of hypothetical trajectories of the dynamic target is determined based on the constant steering rate and acceleration model and the intelligent driver model. If the category of the dynamic target is a vehicle and there are no other vehicles in front of the dynamic target, then the set of assumed trajectories of the dynamic target is determined based on the constant steering rate and acceleration model. If the category of the dynamic target is not a vehicle, then a set of hypothetical trajectories of the dynamic target is determined based on the social force model or the intention-target point model.

[0013] In one or more optional embodiments of this application, determining multiple potential conflict points based on the hypothetical trajectory set of the vehicle and the hypothetical trajectory sets of each dynamic target includes: A spatiotemporal three-dimensional grid is established with the vehicle as the origin; For each data point in the spatiotemporal three-dimensional grid, the vehicle occupancy probability of the data point is obtained based on the set of assumed trajectories of the vehicle. For each data point in the spatiotemporal three-dimensional grid, the target occupancy probability of the data point is obtained based on the set of assumed trajectories of each dynamic target. For each data point in the spatiotemporal three-dimensional grid, if the probability of a vehicle occupying the data point is greater than a first preset threshold and the probability of a target occupying the data point is greater than a second preset threshold, then the data point is identified as a potential conflict point.

[0014] In one or more optional embodiments of this application, constructing a dynamic risk field based on the plurality of potential conflict points includes: An initial dynamic risk field is established with the vehicle as the origin; the initial dynamic risk field is two-dimensional data, and each data point in the initial dynamic risk field corresponds to a position coordinate. Based on each potential conflict point, the severity of the potential conflict point is obtained; Based on the multiple potential conflict points and the severity of each potential conflict point, the risk value of each location coordinate in the initial dynamic risk field is obtained; The dynamic risk field is obtained based on the risk value of each location coordinate.

[0015] Secondly, embodiments of the present invention provide a vehicle prediction and guidance device, which may include: The data acquisition module is used to acquire vehicle status data, driver explicit intent data, environmental perception data, map context data, and navigation path data. The vehicle trajectory prediction module is used to determine a set of hypothetical trajectories for the vehicle based on the vehicle status data, the driver's explicit intent data, the map context data, and the navigation path data. The target trajectory prediction module is used to determine a set of hypothetical trajectories for each dynamic target in the environmental perception data. The first determining module is used to determine multiple potential conflict points based on the hypothetical trajectory set of the vehicle and the hypothetical trajectory set of each dynamic target. A construction module is used to construct a dynamic risk field based on the multiple potential conflict points; The second determining module is used to determine a safe path based on the dynamic risk field; The third determining module is used to determine the guidance instruction based on the predicted collision time and severity of the multiple potential conflict points; The rendering module is used to render the safety path, the dynamic risk field, and the guidance instructions onto the vehicle's A-pillar display screen.

[0016] Thirdly, embodiments of the present invention provide a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the vehicle prediction and guidance method as described above.

[0017] Fourthly, embodiments of the present invention provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the vehicle prediction and guidance method as described above.

[0018] Fifthly, embodiments of the present invention provide a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the vehicle prediction and guidance method as described above.

[0019] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following: This invention provides a vehicle prediction and guidance method. This method generates a set of hypothetical trajectories for vehicles and dynamic targets based on multi-source data, and constructs a dynamic risk field accordingly. It can identify potential conflict points within seconds in advance, elevating safety warnings from traditional real-time responses to proactive prediction, effectively overcoming the warning lag problem of traditional systems. By generating safe paths and guidance instructions based on the dynamic risk field, and accurately presenting the fused guidance information on the vehicle's A-pillar display screen using augmented reality, it provides drivers with intuitive and forward-looking spatial driving suggestions, significantly reducing their cognitive load and decision-making burden in complex scenarios. This achieves a safety paradigm upgrade from passive warning to proactive guidance, significantly improving the prediction and decision-making support capabilities of vehicle safety systems.

[0020] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0021] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0022] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the vehicle prediction and guidance method provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the vehicle prediction and guidance device provided in an embodiment of the present invention. Detailed Implementation

[0023] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0024] The inventors discovered that in existing technologies, mainstream solutions primarily rely on adding sensors (such as cameras and radar) to vehicles to detect obstacles in blind spots and issue warnings to the driver via visual or auditory signals. However, these approaches generally suffer from a fundamental limitation: they operate on a "perception-response" model. That is, the system only alerts or warns the driver when it detects a risky target (such as a vehicle or pedestrian) entering or approaching the vehicle's blind spot. This model has several significant drawbacks in practice: first, warnings are delayed, requiring the target to enter the blind spot or the risk level to be reached before issuing an alarm, leaving the driver with insufficient reaction time; second, cognitive load is high, as most existing technologies directly display raw images or abstract icons, forcing the driver to interpret multiple target information, easily leading to information overload; third, the function is passive, merely acting as a hazard indicator, unable to proactively answer key questions such as whether to turn and how to safely pass based on predictions of the vehicle's intentions and the future behavior of surrounding targets, lacking proactive guidance capabilities. Based on this, the inventors, through further research and development, created this invention, providing a vehicle prediction and guidance method and device.

[0025] Example 1 Embodiment 1 of the present invention provides a vehicle prediction and guidance method, referring to Figure 1 As shown, the method may include the following steps S101-S108: S101: Acquire vehicle status data, driver explicit intent data, environmental perception data, map context data, and navigation path data.

[0026] S102: Determine the set of hypothetical trajectories for the vehicle based on vehicle status data, driver explicit intent data, map context data, and navigation path data.

[0027] S103: For each dynamic target in the environmental perception data, determine the set of hypothetical trajectories for the dynamic target.

[0028] S104: Based on the hypothetical trajectory set of the vehicle and the hypothetical trajectory set of each dynamic target, identify multiple potential conflict points.

[0029] S105: Construct a dynamic risk field based on multiple potential conflict points.

[0030] S106: Determine safe paths based on dynamic risk fields.

[0031] S107: Determine guidance instructions based on the predicted collision time and severity of multiple potential collision points.

[0032] S108: Renders the safe path, dynamic risk field, and guidance instructions onto the vehicle's A-pillar display screen.

[0033] This invention provides a vehicle prediction and guidance method. This method generates a set of hypothetical trajectories for vehicles and dynamic targets based on multi-source data, and constructs a dynamic risk field accordingly. It can identify potential conflict points within seconds in advance, elevating safety warnings from traditional real-time responses to proactive prediction, effectively overcoming the warning lag problem of traditional systems. By generating safe paths and guidance instructions based on the dynamic risk field, and accurately presenting the fused guidance information on the vehicle's A-pillar display screen using augmented reality, it provides drivers with intuitive and forward-looking spatial driving suggestions, significantly reducing their cognitive load and decision-making burden in complex scenarios. This achieves a safety paradigm upgrade from passive warning to proactive guidance, significantly improving the prediction and decision-making support capabilities of vehicle safety systems.

[0034] In step S101 above, vehicle status data, driver explicit intent data, environmental perception data, map context data, and navigation path data are acquired.

[0035] Specifically, vehicle status data can be read in real time via the vehicle controller local area network bus, including vehicle speed, steering wheel angle, yaw rate, longitudinal acceleration, and lateral acceleration.

[0036] Driver explicit intent data is obtained by collecting turn signal switch status and gear position signals, such as the status of left turn signal on, right turn signal on or off, and information on drive, reverse, park, and neutral gears.

[0037] Environmental perception data is collected and fused by cameras and millimeter-wave radar deployed on both sides and rear of the vehicle to form a list containing the position, speed, category, radial speed and azimuth of dynamic targets such as surrounding vehicles, pedestrians and non-motorized vehicles.

[0038] The map context data comes from the high-precision positioning module and the high-precision map, providing the curvature of the lane where the vehicle is located and the lateral offset between the vehicle's centerline and the lane line.

[0039] The navigation route data is obtained from the vehicle's navigation system and is used to indicate the predetermined driving route and the next key action point.

[0040] In step S102 above, a set of hypothetical trajectories for the vehicle is determined based on vehicle status data, driver explicit intent data, map context data, and navigation path data. Specifically, this includes the following steps S1021-S1024: S1021: Short-term trajectory is obtained based on vehicle status data.

[0041] Specifically, this can be achieved by using a vehicle kinematics model, such as a bicycle model, which uses data from vehicle state data, such as vehicle speed and steering wheel angle, as initial conditions to perform recursive calculations and predict the vehicle's trajectory over a short, preset future period. This short-term trajectory reflects the vehicle's physical inertia under current operation and has a high degree of confidence.

[0042] It should be clarified that in this method, the duration of the assumed trajectory is named the preset future duration. The short-term preset future duration in step S1021 refers to the early time window of the preset future duration. For example, the preset future duration can be set to 5 seconds, and the short-term preset future duration can be set to 1.5 seconds.

[0043] S1022: Based on driver explicit intent data, map context data, and navigation path data, obtain driver intent distribution data.

[0044] Specifically, a driving intention probability distribution vector can be constructed as driving intention distribution data. This vector contains probability values ​​corresponding to four driving intentions: lane keeping, changing lanes to the left, changing lanes to the right, and turning. It can be represented as P_I = [p_{keep}, p_{left}, p_{right}, p_{turn}]. Therefore, in this method, the driving intention distribution data is a probabilistic description of the vehicle's potential future high-level driving behaviors (such as going straight, changing lanes, and turning). It is a key intermediate variable connecting the driver's instantaneous actions with the long-term vehicle motion assumptions. For example, when the left turn signal is on and the steering wheel shows a continuous leftward turning tendency, the probability of the "changing lanes to the left" driving intention in the model will be significantly increased.

[0045] Specifically, the probability of each driving intention is calculated using a quantitative model based on multi-source evidence fusion. This model uses the turn signal status from the driver's explicit intention data as strong evidence of the driver's direct driving intention; analyzes the rate of change of steering wheel angle in vehicle state data to quantify the driver's steering urgency; combines the lateral offset of the vehicle's centerline from the lane lines in map context data to determine the degree of vehicle deviation from the lane center; and matches the next key action point provided by the navigation path data with each driving intention as evidence of consistency in long-term path planning. Finally, by assigning preset weights to the above evidence and performing weighted calculations, a quantitative distribution of driving intentions is output, as shown in Formula 1 below:

[0046] In the formula, P(I k ) For driving intent I k The probability, I k ∈{lane keeping, left lane change, right lane change, turning} For driving intent I k The overall weight, This represents the total number of driving intentions.

[0047] Driving intent in Formula 1 above I k The overall weight The calculation formula is shown in Formula 2 below:

[0048] In the formula, d The signal value represents the turn signal status in the driver's explicit intention data. A value of +1 indicates a left turn signal, -1 indicates a right turn signal, and 0 indicates no turn signal. f ( () represents the rate of change of steering wheel angle. The normalization function, g ( d lat) represents the lateral offset of the lane. d The normalization function of lat, 1nav( Ik (Navigation route and driving intent) I k The consistency indicator function (1 for consistency, 0 otherwise). w 1, k , w2 , k , w3 , k and w4 , kFor each piece of evidence regarding driving intent I k The weighting coefficients can be learned from historical data.

[0049] S1023: Based on the driving intention distribution data, determine the long-term hypothetical trajectory set. The driving intention distribution data includes probability values ​​corresponding to multiple driving intentions. Specifically, this includes the following steps S10231-S10234: S10231: Based on the probability value corresponding to each driving intention in the driving intention distribution data, determine multiple possible driving intentions.

[0050] Specifically, one approach is to set a driving intention probability threshold and filter out driving intentions with probability values ​​higher than this threshold from the driving intention distribution data. These are then considered as potential driving intentions for generating specific trajectory hypotheses. For example, if the probabilities of both "lane keeping" and "changing lanes to the left" are higher than the driving intention probability threshold, then these two driving intentions are identified as potential driving intentions that require further processing.

[0051] S10232: Determine the long-term hypothetical trajectory corresponding to each possible driving intention.

[0052] Specifically, for each possible driving intention that passes the screening, a preset trajectory generator that matches it can be invoked.

[0053] For lane-keeping driving intentions, the trajectory generator will take the vehicle's current position and attitude as the starting point, combine the geometry of the current lane centerline and the status of the vehicle in front, and use either a following model or a cruise control model to generate a trajectory that travels along the lane.

[0054] For lane change driving intentions (including left and right lane changes), the trajectory generator will use methods such as fifth-order polynomial curves or optimal control to plan a trajectory that smoothly transitions from the current lane to the target position in the adjacent lane. The duration of the transition process is negatively correlated with the vehicle speed.

[0055] For a turning driving intention, the trajectory generator combines the curvature of the vehicle's lane provided in the map context data to generate a driving trajectory that fits the curve geometry. Ultimately, it generates one or more representative long-term hypothetical trajectories for each possible driving intention.

[0056] S10233: For each long-term hypothetical trajectory, determine the probability value corresponding to the long-term hypothetical trajectory based on the probability value of the driving intention corresponding to the long-term hypothetical trajectory and the dynamic feasibility of the long-term hypothetical trajectory.

[0057] Specifically, the final probability value assigned to a long-term hypothetical trajectory can be determined by two parts: first, the prior probability of the driving intention itself corresponding to the long-term hypothetical trajectory, that is, the probability of that driving intention obtained from the driving intention distribution data. P(I k ) Secondly, there is the kinematic or dynamic feasibility score of the long-term hypothetical trajectory itself, such as the curvature continuity of the trajectory and whether the acceleration exceeds the vehicle's physical limits. By combining the two through weighted summation, a probability value is calculated for the long-term hypothetical trajectory.

[0058] S10234: Combine multiple long-term hypothesis trajectories and the probability value corresponding to each long-term hypothesis trajectory to obtain a set of long-term hypothesis trajectories.

[0059] Specifically, this can be achieved by combining all the long-term hypothetical trajectories generated in step S10232, along with the probability values ​​calculated for each long-term hypothetical trajectory in step S10233, into a set. This set represents all the main driving paths that the vehicle may take within a preset future time period (e.g., 5 seconds) and their probabilities.

[0060] S1024: Based on the short-term trajectory and the long-term hypothetical trajectory set, obtain the vehicle's hypothetical trajectory set. The time length of each hypothetical trajectory in the vehicle's hypothetical trajectory set is a preset future duration.

[0061] Specifically, this can be achieved by temporally connecting and probabilistically fusing short-term trajectories reflecting the near-term certainty of the vehicle with long-term hypothetical trajectories reflecting various possibilities in the future. First, the short-term trajectory generated in step S1021 (e.g., covering the next 1.5 seconds) is used as the initial segment for all complete hypothetical trajectories. Then, for each long-term hypothetical trajectory in the set of long-term hypothetical trajectories, it is smoothly connected to the end of the aforementioned short-term trajectory in terms of position, speed, and heading, thereby forming a complete hypothetical trajectory covering the entire preset future duration (e.g., 5 seconds). Finally, the probability value of this complete hypothetical trajectory is determined by the probability value of the long-term hypothetical trajectory from which it originates.

[0062] This process generates a corresponding complete hypothetical trajectory for each long-term hypothetical trajectory in the long-term hypothetical trajectory set, inheriting its probability value, ultimately forming the vehicle's hypothetical trajectory set. This hypothetical trajectory set can be formally represented as: Ψ{ego}={ (τ{ego, k}, pk) | k = 1, ..., N}, where τ{ego, k} is the vehicle's k-th hypothetical trajectory, ego represents the vehicle itself, pk is the probability value, and N is the total number of hypothetical trajectories. This set comprehensively describes all possible movement paths of the vehicle within a preset future timeframe, taking into account physical inertia, driver intent, and environmental constraints, and their probabilities of occurrence.

[0063] In step S103 above, for each dynamic target in the environmental perception data, a set of hypothetical trajectories for the dynamic target is determined. Specifically, this includes the following steps S1031-S1033: S1031: If the category of the dynamic target is a vehicle and there are other vehicles in front of the dynamic target, then the set of hypothetical trajectories of the dynamic target is determined based on the constant steering rate and acceleration model and the intelligent driver model.

[0064] Specifically, it can be done by first using the Constant Turn Rate and Acceleration (CTRA) model as the basic motion model, and then predicting the basic trajectory of the target vehicle under undisturbed conditions based on its speed, heading, and yaw rate.

[0065] Simultaneously, an Intelligent Driver Model (IDM) is introduced to simulate the vehicle's interaction with the vehicle ahead, thereby correcting the longitudinal motion assumptions in the basic prediction. The IDM calculates the target vehicle's desired acceleration using a continuous function. The core formula of this model is shown in Equation 3 below:

[0066] In the formula, a IDM This represents the expected acceleration calculated by the intelligent driver model. a max It is the maximum acceleration capability of a dynamic target vehicle. v It is the current speed of the dynamic target vehicle. v 0 is the driver's desired free-flow velocity. d It is a preset acceleration index. s It is the actual distance between the dynamic target vehicle and the vehicle in front. s*(v,Δv) This is the expected safe distance between vehicles.

[0067] The desired safe vehicle spacing in Formula 3 above s*(v,Δv) The calculation method is shown in Formula 4 below:

[0068] In the formula, s 0 is the minimum safe distance when stationary, and T is the expected safe headway. v It is the current speed of the dynamic target vehicle. Δv It is the relative speed between the dynamic target vehicle and the vehicle in front (speed of the vehicle in front minus speed of the dynamic target vehicle). a max It is the maximum acceleration capability of a dynamic target vehicle. bcomf is for comfortable deceleration.

[0069] In application, the current expected safe distance between the target vehicle and the vehicle in front is first calculated based on the real-time status of the dynamic target vehicle and the vehicle in front, using the above formula 4. s*(v,Δv) Then, compare this safe distance with the actual distance between vehicles. s Substituting these values ​​into Formula 3 above, we can calculate the expected acceleration that the driver of the dynamic target vehicle is most likely to take in this interactive scenario. a IDM Finally, the calculated expected acceleration value is used to correct or replace the "constant acceleration" parameter assumed in the original constant steering ratio and acceleration model, thereby generating a set of longitudinal motion trajectories that better fit the logic of real-world following vehicle interaction. Combined with the model's original lateral motion (constant steering ratio) prediction, a complete set of assumed trajectories for the dynamic target vehicle in the interaction scenario is finally formed.

[0070] S1032: If the category of the dynamic target is a vehicle and there are no other vehicles in front of the dynamic target, then the set of assumed trajectories of the dynamic target is determined based on the constant steering rate and acceleration model.

[0071] Specifically, this can be achieved by directly applying the constant steering rate and acceleration model described in step S1031 above. Based on the current motion state vector of the dynamic target vehicle (including position, velocity, heading angle, steering rate, and acceleration), assuming that it maintains its current steering rate and acceleration unchanged during the prediction period, a set or more kinematically coherent trajectories within a preset future time period are generated through model recursion, forming its hypothetical trajectory set. In this case, the target's motion prediction does not involve complex interactions with other vehicles.

[0072] S1033: If the category of the dynamic target is not a vehicle, then determine the set of hypothetical trajectories of the dynamic target based on the social force model or the intention-target point model.

[0073] Specifically, based on the accuracy of available sensor data, the richness of scene semantic information, and computational resources, an appropriate prediction model can be selected, such as a Social Force Model or an Intention-Destination Model. When using a Social Force Model, the target is considered a particle in a virtual force field composed of attractive forces (e.g., the target point), repulsive forces (e.g., obstacles, vehicles, other pedestrians), and boundary forces (e.g., the edge of a sidewalk). By calculating the forces acting on it, its future movement path is simulated and predicted, generating a set of probabilistic hypothetical trajectories, resulting in a set of hypothetical trajectories. When using an Intention-Destination Model, the target's behavior (e.g., pausing and looking around, changing direction) and its surrounding scene (e.g., intersections, crosswalks, bus stops) are first analyzed to infer its possible intentions and destinations. Subsequently, based on the inferred intentions, one or more reasonable paths are planned for it (e.g., crossing the street at a crosswalk). Finally, its future position sequence is generated along these paths, forming a set of hypothetical trajectories. Regardless of the model used, the prediction process can significantly incorporate the influence of scene semantic information such as crosswalks, intersections, and stationary obstacles.

[0074] Through the above steps S1031-S1033, the set of hypothetical trajectories for all dynamic targets is obtained, which can be formally represented as: Θj = {τ{target, j, m} | m = 1, ..., Mj}, where τ{target, j, m} is the m-th hypothetical trajectory of the j-th dynamic target, target represents the dynamic target, and Mj is the total number of hypothetical trajectories of the j-th dynamic target.

[0075] In step S104 above, multiple potential conflict points are determined based on the hypothetical trajectory set of the vehicle and the hypothetical trajectory set of each dynamic target. Specifically, this includes the following steps S1041-S1044: S1041: Establish a spatiotemporal three-dimensional grid with the vehicle as the origin.

[0076] Specifically, this can be achieved by defining a discretized three-dimensional coordinate system, i.e., a spatiotemporal three-dimensional grid, with the vehicle as the origin, covering a preset future time period and a certain spatial range around the vehicle. The three dimensions of this spatiotemporal three-dimensional grid are time, spatial vertical dimension, and spatial horizontal dimension. Time and spatial range are discretized according to a preset resolution, thus forming a spatiotemporal three-dimensional grid G(t, x, y) composed of numerous data points. Each data point represents a spatiotemporal unit at a specific time and location in the future.

[0077] S1042: For each data point in the spatiotemporal three-dimensional grid, based on the set of assumed vehicle trajectories, obtain the vehicle occupancy probability of the data point.

[0078] Specifically, this could involve iterating through each hypothetical trajectory (τ{ego,k}, pk) in the set of hypothetical trajectories Ψ{ego} for the vehicle. For a given hypothetical trajectory τ{ego, k}, calculate the spatial location it will pass through or be closest to at each discrete time point in the future. For a specific data point G(t) in the spatiotemporal three-dimensional raster... i , x j , y l ), examine the trajectory τ{ego,k} at time t i Has it been passed by (x) j , y l The hypothetical trajectory is defined as a spatial region centered on a given point and within a certain tolerance range. If the trajectory passes through this region, it is considered that the hypothetical trajectory "occupies" this data point.

[0079] The vehicle at this data point G(t) i , x j , y l The probability of a vehicle occupying a point is the sum of the probabilities pk of all hypothetical trajectories that "occupy" that data point.

[0080] S1043: For each data point in the spatiotemporal three-dimensional grid, based on the set of assumed trajectories of each dynamic target, obtain the target occupancy probability of the data point.

[0081] Specifically, for each dynamic target j, it has a multimodal set of hypothetical trajectories Θj = {τ{j,m}}, which also includes w{j,m}, representing the likelihood weight of the hypothetical trajectory mode τ{j,m}, obtained in step S103 above.

[0082] For a specific data point G(t) in the spatiotemporal raster i , x j , y l First, calculate the weight value of a single dynamic target j at that data point: iterate through all hypothetical trajectories in Θj, and assign the weight values ​​of those trajectories at time t. i Passing through point (x) j , y l The weights w{j,m} of the hypothetical trajectory in the vicinity of the target j are summed to obtain the occupancy weight value of the dynamic target j. Then, this process is repeated for all dynamic targets j, and the occupancy weight values ​​calculated for each dynamic target are summed again. The final sum is the target occupancy probability of the data point.

[0083] S1044: For each data point in the spatiotemporal three-dimensional grid, if the probability of a vehicle occupying the data point is greater than the first preset threshold and the probability of a target occupying the data point is greater than the second preset threshold, then the data point is identified as a potential conflict point.

[0084] Specifically, it could be to iterate through each data point G(t) in the spatiotemporal three-dimensional raster. i , x j , y l Simultaneously check the vehicle occupancy probability and the target occupancy probability. If the conditions are met: the vehicle occupancy probability is greater than a first preset threshold and the target occupancy probability is greater than a second preset threshold, then it is considered that at this future time t... i In spatial location (x) j , y l At point ), there is a significant possibility of collision between the vehicle and at least one dynamic target.

[0085] Therefore, this data point is marked as a potential conflict point (CP). n And record its spatiotemporal coordinates (t) n , x n , y n All marked points constitute a list of potential conflict points {CP}. n}. Wherein, the time coordinate t of the conflict point. n This refers to predicting the collision time. The selection of the first and second preset thresholds is used to balance the system's sensitivity and false alarm rate.

[0086] In step S105 above, a dynamic risk field is constructed based on multiple potential conflict points. Specifically, this includes the following steps S1051-S1054: S1051: Establish an initial dynamic risk field with the vehicle as the origin. The initial dynamic risk field is two-dimensional data, and each data point in the initial dynamic risk field corresponds to a position coordinate.

[0087] Specifically, the two-dimensional spatial range covered by the initial dynamic risk field can be consistent with the spatial range (i.e., x and y dimensions) of the spatiotemporal three-dimensional grid established in step S1041. For example, if the spatiotemporal three-dimensional grid covers a rectangular area centered on the vehicle and extending tens of meters in all directions, and is discretized with a certain spatial resolution (e.g., 0.5 meters), then the initial dynamic risk field established in this step will define a two-dimensional grid with the exact same spatial coverage and resolution. Each cell in this grid, i.e., each data point, is uniquely identified by its coordinates (x, y), corresponding to a planar position in the real world, and also corresponding one-to-one with the spatial position of the spatiotemporal three-dimensional grid slice at the same time. Initially, the risk value of all position coordinates in this two-dimensional grid is preset to zero or a basic value. This two-dimensional dynamic risk field essentially maps the results of the above spatiotemporal analysis (potential conflict points) to a two-dimensional continuous representation that facilitates spatial path planning and risk visualization. The consistency of its spatial definition ensures logical coherence and lossless data conversion from conflict detection to the construction of the dynamic risk field.

[0088] S1052: Based on each potential conflict point, obtain the severity of the potential conflict point.

[0089] Specifically, this could be for each potential conflict point CP identified in step S104. n A severity value is calculated based on the associated predicted collision information. n This severity score comprehensively reflects the potential severity of the consequences should a collision occur. The calculation primarily considers three factors: first, the relative velocity at the potential point of impact; a higher relative velocity indicates higher collision energy and thus greater severity; second, the weighting coefficients corresponding to the target types involved at the potential point of impact, which can be exemplarily set as pedestrian > bicycle > motorcycle > car > truck, reflecting vulnerability to injury; and finally, the predicted collision angle, as side collisions are generally more dangerous than rear-end collisions. The formula for calculating the severity of a potential point of impact is shown in Formula 5 below:

[0090] In the formula, Severity n The severity of the nth potential conflict point. v rel, n The relative velocity at the potential point of conflict. v max This is a preset reference speed or maximum normalized speed used to normalize relative speeds. or (type j ) represents the weighting coefficient corresponding to the target type involved in the potential conflict point. i collision n The collision angle.

[0091] S1053: Based on multiple potential conflict points and the severity of each potential conflict point, obtain the risk value of each location coordinate in the initial dynamic risk field.

[0092] Specifically, for each location coordinate (x, y) in the initial dynamic risk field, its risk value R(x,y) is the sum of the influences of all potential conflict points on that location coordinate. Each potential conflict point CP n The risk contribution of location coordinates (x,y) is related to its own severity. n Proportional to, and increasing with, the spatial location (x) of the potential conflict point from that location coordinate. n , y n The distance between the current time t0 and the predicted collision time t from the potential conflict point. n (i.e., TTC) nThe risk value decays due to the time difference. Specifically, the risk value at location coordinates (x, y) can be calculated using the following formula 6:

[0093] In the formula, R ( x, y ) represents the risk value at location coordinates (x, y). N CP Severity is the total number of potential conflict points. n The severity of the nth potential conflict point. x n and y n Indicates the location of the potential conflict point in the vehicle coordinate system, TTC n The predicted collision time from the current time t0 to the nth potential conflict point. ss and s This is a spatial and temporal attenuation coefficient used to control the spread of risk impact. The formula implies that the risk impact of a potential conflict point spreads in a "cloud" shape in space and time; the closer to the conflict point and the closer to the moment of collision, the greater the impact.

[0094] S1054: Based on the risk value of each location coordinate, a dynamic risk field is obtained.

[0095] Specifically, after calculating the risk values ​​of all location coordinates in the dynamic risk field in step S1053, a complete and numerical dynamic risk field is obtained. This dynamic risk field R(x, y) is a scalar field defined in a two-dimensional space around the vehicle. The value of each point in the field represents the comprehensive collision risk probability of that location within a preset future timeframe. This field transforms the discrete list of potential conflict points generated based on probability trajectory prediction into a continuous, spatialized risk distribution map, providing direct input for subsequent safe path planning and risk visualization. This dynamic risk field is periodically recalculated and updated (e.g., 10 times per second) as the vehicle and target move and the prediction is updated.

[0096] In step S106 above, a safe path is determined based on the dynamic risk field.

[0097] Specifically, the dynamic risk field can be viewed as a cost map, where the risk value R(x, y) for each location coordinate (x, y) represents the "cost" or "price" of traversing that location. The problem of planning a safe path is transformed into finding a safe path on this dynamic risk field that starts from the vehicle's current location and extends to a predetermined future time, minimizing the overall cost of this safe path. To achieve this, a cost function is defined to evaluate any candidate path. This cost function weighs the path's safety, comfort, and degree of adherence to the driver's original intent, and its calculation formula is shown in Equation 7 below:

[0098] In the formula, Cost( t The cost of candidate paths is ). T h To preset future duration, t ( t ) represents a candidate path t At time t, in the spatial position R ( t ( t )) indicates the location of the dynamic risk field. t ( t The risk value at point ) k ( t ) indicates the path at time t curvature, t intent( t This represents a reference trajectory that reflects the driver's original intention (e.g., the desired trajectory generated based on a simple turn signal). α , β , c These are preset weighting coefficients, used to balance the three optimization objectives of risk avoidance, ride comfort, and intention following.

[0099] Subsequently, the dynamic risk field is thresholded, defining areas with risk values ​​below a preset safety threshold as passable low-risk areas. Then, within these low-risk areas, a path search algorithm (e.g., A* algorithm, fast random tree, or its variant) is used for planning. By optimizing the aforementioned cost function, a path is found that minimizes the cost (…). t The algorithm ultimately outputs the shortest path. This continuous path, starting from the vehicle's current position and extending over a preset future time, is the safe path. Spatially, it avoids high-risk areas; kinematically, it meets vehicle dynamics constraints and comfort requirements; and it closely matches the driver's steering intentions, providing the geometric basis for subsequent guidance information generation. This safe path is periodically replanned as the dynamic risk field is updated.

[0100] In step S107 above, guidance instructions are determined based on the predicted collision times and severity of multiple potential conflict points. Specifically, this includes the following steps S1071-S1072: S1071: For each potential conflict point, if the predicted collision time is less than the first preset warning time or the severity is greater than the preset severity threshold, a strong intervention guidance instruction is generated. The strong intervention guidance instruction includes avoidance suggestions.

[0101] Specifically, this can involve iterating through all potential conflict points and checking the predicted collision time and severity for each. If a potential conflict point meets the condition that its predicted collision time is less than a first preset warning time or its severity is greater than a preset severity threshold, then the risk is deemed imminent and the consequences serious, requiring a clear and strong intervention instruction to be issued to the driver. The strong intervention guidance instruction generated at this time directly includes specific evasive action suggestions in its semantic content, such as "Do not change lanes to the left," "Emergency braking recommended," or "Please immediately make a slight right turn." This type of instruction aims to guide the driver to perform specific evasive actions in the clearest and most unambiguous way. The first preset warning time can be set to 3 seconds, for example.

[0102] S1072: If the predicted collision time at the potential conflict point is greater than or equal to the first preset warning time and less than the second preset warning time, a suggested guidance instruction is generated. The suggested guidance instruction includes a recommended safe driving area. The first preset warning time is less than the second preset warning time.

[0103] Specifically, if a potential conflict point meets the condition that the predicted collision time is greater than or equal to the first preset warning time and less than the second preset warning time, then the risk is determined to exist in the medium to future, and there is still relatively ample time for decision-making and action. The generated advisory guidance instructions do not specify concrete evasive actions, but rather focus on alerting the existence of the risk and indicating safe driving space. For example, the instructions might be "A vehicle is rapidly approaching on the left; please be aware of the space on the right" or "There is a risk at the merging area ahead; it is recommended to drive in the center of the current lane." These instructions aim to enhance the driver's risk awareness and guide their attention to safe alternative areas, allowing the driver to make their own decision on the specific action. The second preset warning time can be set to 5 seconds, for example.

[0104] In this embodiment, potential conflict points with a predicted collision time greater than or equal to a second preset warning time are determined to be long-term risks with low current urgency. In this case, no specific avoidance or suggested action instructions are generated; instead, the process proceeds to the information display level. Specifically, this may involve visually marking or subtly highlighting the risk area associated with such conflict points (through dynamic risk field mapping), for example, displaying it as a semi-transparent, static, or slowly changing risk cloud outline on the A-pillar display screen, without any directional arrows, prohibition symbols, or text suggestions. This level of processing aims to provide the driver with complete situational awareness, enabling them to understand potential long-term risks, but avoiding unnecessary interference before the risk is imminent, leaving the decision-making power entirely to the driver, and acting only as an auxiliary information provider.

[0105] In step S108 above, the safe path, dynamic risk field, and guidance instructions are rendered onto the vehicle's A-pillar display screen. Specifically, this includes the following steps S1081-S1082: S1081: Convert the security path, dynamic risk field, and boot instructions into a boot image. This specifically includes the following steps: S10811-S10814: S10811: Render the safety path as a first guide graphic. The first guide graphic is a semi-transparent filled strip area with a preset safety color. Its transparency is determined according to the predicted collision time of potential conflict points in the safety path, and its edges have a dynamic flowing light effect to indicate the driving direction.

[0106] Specifically, it can be based on the centerline of the safe path obtained in step S106, extending outwards to both sides according to the vehicle width or a preset safety margin to form a strip-shaped polygonal region with width. This region is then semi-transparently filled with a preset safety color (such as green) as the first guiding graphic. Its transparency is not fixed, but rather determined based on the predicted collision time (TTC) of the most pressing potential conflict point associated with the safe path. closest Dynamic adjustment. The formula for the transparency of the first guiding graphic is as follows:

[0107] In the formula, α channel ( t The first guiding graphic at the current moment is... t transparency, α base Based on basic transparency (e.g., 0.3), TTC closest The predicted collision time for the most imminent potential conflict point. T h This is a preset future duration.

[0108] This formula ensures that when TTC...closest As the current time approaches, the transparency approaches the base transparency; lower transparency makes the graphic more prominent. (TTC) closest As the distance from the current moment increases, the transparency approaches 1; higher transparency results in a fainter graphic. This achieves an adaptive match between the intensity of visual guidance and the urgency of the risk. Simultaneously, at the edge of the strip-shaped area of ​​the first guiding graphic, a dynamic flowing light effect is rendered along the path's extension direction. For example, brighter, same-color light spots or bands move along the edge, visually indicating the recommended driving direction.

[0109] S10812: Render high-risk areas in the dynamic risk field with risk values ​​exceeding a preset risk threshold as a second guiding graphic. The second guiding graphic is a semi-transparent cloud-shaped or mesh-like area filled with a preset warning color. Its transparency is determined according to the risk value of the high-risk area, and it is accompanied by a periodic pulsating warning effect.

[0110] Specifically, this can be achieved by applying a preset risk threshold to the dynamic risk field obtained in step S105. All continuous spatial regions within the field with risk values ​​exceeding the preset risk threshold are extracted and designated as high-risk areas. The outlines of these areas are delineated using a preset warning color (such as red or orange) and filled with semi-transparent material to form a cloud-like shape, serving as the second guide graphic. The transparency of the second guide graphic is positively correlated with the risk value at the corresponding position coordinates; the higher the risk value, the lower the transparency. The formula for the transparency of the second guide graphic is as follows:

[0111] In the formula, α risk x,y The second guiding graphic is located at coordinates ( x,y) transparency, α base Based on basic transparency (e.g., 0.3). R ( x,y ) is the position coordinate ( x,y The risk value of ) R max and R min is the normalized boundary of the dynamic risk field, which can be a fixed value or dynamically calculated based on the current field conditions.

[0112] This formula linearly maps the risk value to a transparency range. α base [1] Above. In addition, to make the warning effect more prominent, the overall brightness or transparency of the graphic will pulsate periodically at a certain frequency, that is, the intensity will change alternately, thereby dynamically attracting the driver's attention.

[0113] S10813: Render the boot instructions as a third boot graphic.

[0114] Specifically, based on the semantic type and urgency of the guidance instructions generated in step S107, they can be converted into corresponding standardized graphic symbols or concise text labels. For example, for strong intervention guidance instructions such as "No lane change," a red prohibition symbol can be rendered overlaid on the corresponding lane position on the A-pillar display screen on the corresponding side; for "Recommended braking" instructions, a flashing brake icon can be rendered; for advisory instructions such as "Pay attention to right-hand space," an arrow outline pointing to a safe area can be rendered on the right side of the screen. The specific style, color, and dynamic effects (such as flashing) of the third guidance graphic are strictly bound to the level of the guidance instruction (strong intervention, advisory).

[0115] S10814: Based on the first guide graphic, the second guide graphic, and the third guide graphic, a guide image is obtained.

[0116] Specifically, the first, second, and third guide graphics generated in steps S10811, S10812, and S10813, respectively, can be superimposed onto the same image layer or graphics buffer according to their respective spatial coordinates. During this process, potential occlusion and superimposition effects need to be addressed based on the spatial relationships of the graphics, for example, ensuring that the first guide graphic (safety passage) is always displayed on top or has a clear visual distinction from the second guide graphic (risk area). Ultimately, this composite layer containing all the guide elements becomes the guide image to be displayed on the A-pillar display screen.

[0117] S1082: By using spatial registration and perspective fusion technology, the guide image is displayed on the vehicle's A-pillar display screen, so that the guide image is visually spatially aligned with the real external scene observed by the vehicle driver through the A-pillar.

[0118] Specifically, the A-pillar display screen can be calibrated in advance during vehicle production or safety system initialization. This calibration process establishes a precise mapping between the display screen's pixel coordinate system and the vehicle's fixed coordinate system, taking into account the driver's eye position. When the vehicle is running, each graphic element in the guidance image (such as a point on the first guidance image or an area on the second guidance image) is associated with a coordinate position in the actual three-dimensional world. For example, a point on a safe path corresponds to an (X, Y) position in the world coordinate system, and the center of a high-risk area also corresponds to a world coordinate. Using the predefined calibration mapping relationship and the current vehicle attitude (such as yaw angle), the specific pixel coordinates that the three-dimensional world coordinate point should be projected onto the A-pillar display screen are calculated in real time. Subsequently, the guidance image rendering engine draws the graphic elements onto the corresponding positions on the display screen based on these calculated pixel coordinates.

[0119] Because this coordinate transformation process strictly follows the principles of geometric optical perspective and compensates for factors such as the display screen's installation position and the driver's viewing angle, the virtual guidance graphics presented on the display screen (such as the green safety strip area and the red risk cloud) are precisely aligned or closely related in spatial position to the corresponding roads, vehicles, pedestrians, and other objects in the real world seen through the transparent A-pillar (or the transparent display area of ​​the A-pillar) as perceived by the driver. This achieves an augmented reality-style fusion of virtual information and the real environment. This process ensures that the guidance information is presented intuitively and without cognitive bias next to its indicated real spatial location, completing the crucial transformation from abstract data to intuitive spatial guidance.

[0120] In this embodiment, while the guidance information is presented through the A-pillar display screen, adaptive display and human-machine interaction logic are also included to ensure that the guidance information is clear and effective in various environments and does not interfere with the driver. Specifically, this includes three aspects: ambient light adaptation, driver attention coupling, and multimodal alarm synchronization. First, to achieve adaptive ambient light adjustment, the A-pillar display screen has an integrated or associated ambient light sensor that detects the light intensity in the cockpit in real time. Based on this light intensity, the overall brightness of the A-pillar display screen and the overall contrast of the guidance graphics are dynamically adjusted to ensure that the guidance information is clearly visible and glare-free under any lighting conditions (such as strong daylight, nighttime, and tunnel entry / exit). The formula for adjusting the overall brightness of the A-pillar display screen is as follows:

[0121] In the formula, B screen refers to the global brightness of the A-pillar display screen. B min and B `max` represents the minimum and maximum global brightness (e.g., 0.1 and 1.0). L env represents the ambient light intensity. L 0 represents the baseline illuminance, such as 1000 lux, while Lmax represents the maximum design illuminance, such as 100,000 lux, corresponding to direct sunlight.

[0122] The formula for adjusting the overall contrast of the graphic is shown below:

[0123] In the formula, C To guide the overall contrast of the graphic. C min and C `max` represents the minimum and maximum contrast ratio. C base As the baseline contrast, L env represents the ambient light intensity. L 0 represents the baseline illumination intensity.

[0124] Through the above adaptive adjustments, the optimal readability and visual comfort of the visual guidance interface under different ambient light conditions are ensured.

[0125] To achieve driver attention coupling, a quantification algorithm is used to precisely adjust the alarm intensity. When the driver's gaze is not focused on the guidance or a high-risk area, the salience of the multimodal alarm is dynamically increased based on the duration of gaze absence. Specifically, an adjustment factor is calculated using an alarm intensity gain formula. The formula is as follows:

[0126] In the formula, A Boost is the alarm intensity gain factor, which will be applied to parameters such as the current visual flicker frequency, audible volume, and tactile vibration intensity. k Δ is the preset gain coefficient (e.g., set to 1.0). t gaze is the duration for which the driver's line of sight last leaves the target guidance area or risk area, Δ t threshold is a preset threshold for the time of attention loss (e.g., 1.5 seconds).

[0127] Formula 12 above shows that the time Δ when the driver's gaze leaves the target area t The closer gaze is to the threshold Δ t threshold, the value within parentheses approaches 0, gain factor. A The boost value approaches 1, meaning no boost is applied. This mechanism ensures that when the driver may be distracted, their attention is proactively and smoothly regained with an intensity appropriate to the duration of the distraction and the urgency of the risk, thereby ensuring that critical safety information is effectively communicated when necessary.

[0128] To achieve multimodal alarm synchronization, this method's guidance and warning do not rely solely on the visual channel. The warning levels of the visual guidance graphics (e.g., constant light, slow flashing, fast flashing) are strictly synchronized with the vehicle's existing auditory alarms (e.g., buzzer tone, frequency, and voice prompts) and tactile alarms (e.g., vibration patterns and intensity of the steering wheel or seat), and are controlled by a unified guidance decision logic. For example, when a "strong intervention guidance command" is generated, the corresponding third guidance graphic (e.g., a prohibition icon) on the A-pillar display will flash rapidly, while the steering wheel will emit a strong vibration at a synchronized frequency, accompanied by a specific warning sound. This consistency in timing, rhythm, and semantics of cross-modal signals constructs a unified and strong warning system that can more quickly and reliably awaken the driver's attention and convey clear action intentions, thereby shortening the driver's reaction time and improving the effectiveness of evasive maneuvers in emergency situations.

[0129] This embodiment also includes a real-time update and smooth exit mechanism for guidance information. The guidance information is updated cyclically at a frequency of no less than 10Hz. Based on the latest trajectory prediction and dynamic risk field calculation results, the shape, position, color, and dynamic effects of the guidance graphics are dynamically adjusted to ensure they remain synchronized with the rapidly changing traffic situation. When the predicted risk is eliminated (e.g., the overall dynamic risk field is below a safe threshold) or the driver's intention changes (e.g., turn signals are turned off, steering wheel is straightened), making guidance unnecessary, the A-pillar display does not immediately turn off. Instead, the guidance graphics (e.g., safety lanes, risk area outlines) fade out smoothly within approximately 1 second, avoiding driver confusion due to sudden image disappearance. When the vehicle completely leaves the current triggering scenario (e.g., completing a lane change, leaving an intersection), the A-pillar display will return to a transparent display or low-power standby state after the guidance graphics fade out, awaiting activation of the next scenario. This mechanism ensures the continuity and naturalness of human-computer interaction, improving the overall user experience.

[0130] In summary, the beneficial effect of this method is that it achieves significant progress in the field of vehicle blind spot safety through a complete technical solution.

[0131] First, this method represents a fundamental shift in the safety paradigm, moving from "passive, delayed early warning" to "proactive, forward-looking guidance." Traditional methods suffer from warning lag due to reliance on targets entering blind spots. This method, by determining a set of hypothetical trajectories for vehicles and dynamic targets based on multi-source data and constructing a dynamic risk field, can proactively identify potential conflict points within seconds. The construction of the dynamic risk field advances safety assessment from the current physical space to the future spatiotemporal domain, enabling the generated safe paths and guidance instructions to provide drivers with a valuable window of decision-making and action time before the actual occurrence of danger, thereby reducing accident risk at its source.

[0132] Second, this method constructs a tightly coupled "prediction-decision-guidance" complete technical chain, forming a high technical barrier. This chain consists of probabilistic trajectory prediction, spatiotemporal conflict quantification, and dynamic risk field visualization mapping, forming an end-to-end organic whole. Specifically, the set of hypothetical trajectories of vehicles and dynamic targets provides a probabilistic description of the future scenario. The dynamic risk field quantifies this description into a continuous spatial risk distribution. Based on the dynamic risk field and potential conflict points, safe paths and guidance instructions are determined, and finally, spatial registration and perspective fusion techniques are used to transform them into intuitive guidance images. This complete chain is the core technical foundation for generating predictive guidance effects.

[0133] Third, this method provides "augmented reality" guidance that aligns with human spatial cognitive intuition, significantly reducing the driver's cognitive load. Through spatial registration and perspective fusion technology, guidance graphics such as safe routes and high-risk areas in dynamic risk fields are precisely overlaid and displayed on the A-pillar display screen at the corresponding real-world location, achieving visual alignment between virtual information and the real-world scene. This "what you see is what you are guided" approach allows drivers to understand risky locations and avoidance routes in a near-intuitive way, significantly shortening information comprehension and decision-making time.

[0134] Fourth, this method deepens the utilization of vehicle status signals, achieving intelligent adaptation across all scenarios. This method goes beyond simple triggering in its use of vehicle status data and driver explicit intent data, achieving a deep understanding of complex driving scenarios. Simultaneously, through ambient light adaptive adjustment and driver attention coupling mechanisms, it can dynamically optimize display and alarm methods based on external lighting and driver status, ensuring effectiveness and comfort under various complex operating conditions.

[0135] Fifth, this method enhances reliability and robustness, providing crucial safety redundancy for advanced autonomous driving. As an independent prediction-based active safety layer, its core algorithm, based on a probabilistic framework and multiple hypothesis analysis, exhibits stronger fault tolerance to sensor noise or target misidentification. Its output safe path can serve as a reference input or cross-validation basis for the vehicle planning and control module, thereby improving the overall system's decision-making safety.

[0136] Example 2 Based on the same inventive concept, embodiments of the present invention also provide a vehicle prediction and guidance device, referring to... Figure 2 As shown, the device includes: The data acquisition module 101 is used to acquire vehicle status data, driver explicit intent data, environmental perception data, map context data, and navigation path data; The vehicle trajectory prediction module 102 is used to determine a set of hypothetical trajectories for the vehicle based on the vehicle status data, the driver explicit intent data, the map context data, and the navigation path data. The target trajectory prediction module 103 is used to determine a set of hypothetical trajectories for each dynamic target in the environmental perception data. The first determining module 104 is used to determine multiple potential conflict points based on the hypothetical trajectory set of the vehicle and the hypothetical trajectory set of each dynamic target. Construction module 105 is used to construct a dynamic risk field based on the multiple potential conflict points; The second determining module 106 is used to determine a safe path based on the dynamic risk field; The third determining module 107 is used to determine the guidance instruction based on the predicted collision time and severity of the multiple potential conflict points; The rendering module 108 is used to render the safety path, the dynamic risk field, and the guidance instructions onto the vehicle's A-pillar display screen.

[0137] Example 3 Based on the same inventive concept, embodiments of the present invention also provide a computer-readable storage medium storing a computer program / instructions thereon, which, when executed by a processor, implements the vehicle prediction and guidance method as described in Embodiment 1 above.

[0138] Example 4 Based on the same inventive concept, embodiments of the present invention also provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the vehicle prediction and guidance method as described in Embodiment 1 above.

[0139] Example 5 Based on the same inventive concept, this embodiment of the invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory. When the processor executes the computer program, it implements the vehicle prediction and guidance method as described in Embodiment 1 above.

[0140] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0141] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure one One or more processes and / or boxes Figure one A device that provides the functions specified in one or more boxes.

[0142] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure one One or more processes and / or boxes Figure one The function specified in one or more boxes.

[0143] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure one One or more processes and / or boxes Figure one The steps of the function specified in one or more boxes.

[0144] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A vehicle prediction and guidance method, characterized in that, include: Acquire vehicle status data, driver explicit intent data, environmental perception data, map context data, and navigation path data; Based on the vehicle status data, the driver's explicit intent data, the map context data, and the navigation path data, a set of hypothetical trajectories for the vehicle is determined; For each dynamic target in the environmental perception data, determine a set of hypothetical trajectories for the dynamic target; Based on the hypothetical trajectory set of the vehicle and the hypothetical trajectory set of each dynamic target, multiple potential conflict points are identified. A dynamic risk field is constructed based on the aforementioned multiple potential conflict points; Based on the dynamic risk field, a safe path is determined; Based on the predicted collision times and severity of the multiple potential conflict points, guidance instructions are determined. The safe path, the dynamic risk field, and the guidance instructions are rendered onto the vehicle's A-pillar display screen.

2. The method according to claim 1, characterized in that, The step of rendering the safety path, the dynamic risk field, and the guidance instructions onto the vehicle's A-pillar display screen includes: The safe path, the dynamic risk field, and the guidance instructions are converted into a guidance image; By using spatial registration and perspective fusion technology, the guide image is displayed on the vehicle's A-pillar display screen, so that the guide image is visually spatially aligned with the real external scene observed by the vehicle driver through the A-pillar.

3. The method according to claim 2, characterized in that, The step of converting the security path, the dynamic risk field, and the guidance instructions into a guidance image includes: The safe path is rendered as a first guide graphic; the first guide graphic is a semi-transparent filled strip area with a preset safety color, the transparency of which is determined according to the predicted collision time of potential conflict points in the safe path, and the edges are provided with dynamic flowing light effects to indicate the driving direction. The high-risk area in the dynamic risk field with a risk value higher than a preset risk threshold is rendered as a second guide graphic; the second guide graphic is a semi-transparent cloud-shaped or mesh-shaped area filled with a preset warning color, the transparency of which is determined according to the risk value of the high-risk area, and is accompanied by a periodic pulsating warning effect. The boot instructions are rendered as a third boot graphic; The guide image is obtained based on the first guide graphic, the second guide graphic, and the third guide graphic.

4. The method according to claim 1, characterized in that, The process of determining guidance instructions based on the predicted collision times and severity of the multiple potential conflict points includes: For each potential conflict point, if the predicted collision time of the potential conflict point is less than a first preset warning time or the severity is greater than a preset severity threshold, a strong intervention guidance instruction is generated; the strong intervention guidance instruction includes avoidance suggestions. If the predicted collision time of the potential conflict point is greater than or equal to the first preset warning time and less than the second preset warning time, a suggested guidance instruction is generated; the suggested guidance instruction includes a recommended safe driving area; the first preset warning time is less than the second preset warning time.

5. The method according to claim 1, characterized in that, The step of determining the set of hypothetical trajectories for the vehicle based on the vehicle status data, the driver's explicit intent data, the map context data, and the navigation path data includes: Based on the vehicle status data, a short-term trajectory is obtained; Based on the driver's explicit intent data, the map context data, and the navigation path data, driving intent distribution data is obtained; Based on the driving intention distribution data, a set of long-term hypothetical trajectories is determined; Based on the short-term trajectory and the long-term hypothetical trajectory set, the hypothetical trajectory set of the vehicle is obtained; the time length of each hypothetical trajectory in the hypothetical trajectory set of the vehicle is a preset future duration.

6. The method according to claim 5, characterized in that, The driving intention distribution data includes probability values ​​corresponding to multiple driving intentions; The determination of the long-term hypothetical trajectory set based on the driving intention distribution data includes: Based on the probability value corresponding to each driving intention in the driving intention distribution data, multiple possible driving intentions are determined; Determine the long-term hypothetical trajectory corresponding to each possible driving intention; For each of the long-term hypothetical trajectories, the probability value corresponding to the long-term hypothetical trajectory is determined based on the probability value of the driving intention corresponding to the long-term hypothetical trajectory and the dynamic feasibility of the long-term hypothetical trajectory. By combining multiple long-term hypothesis trajectories and the probability value corresponding to each long-term hypothesis trajectory, the set of long-term hypothesis trajectories is obtained.

7. The method according to claim 1, characterized in that, For each dynamic target in the environmental perception data, determining the set of hypothetical trajectories for the dynamic target includes: If the category of the dynamic target is a vehicle and there are other vehicles in front of the dynamic target, then the set of hypothetical trajectories of the dynamic target is determined based on the constant steering rate and acceleration model and the intelligent driver model. If the category of the dynamic target is a vehicle and there are no other vehicles in front of the dynamic target, then the set of assumed trajectories of the dynamic target is determined based on the constant steering rate and acceleration model. If the category of the dynamic target is not a vehicle, then a set of hypothetical trajectories of the dynamic target is determined based on the social force model or the intention-target point model.

8. The method according to claim 1, characterized in that, Based on the hypothetical trajectory set of the vehicle and the hypothetical trajectory set of each dynamic target, multiple potential conflict points are identified, including: A spatiotemporal three-dimensional grid is established with the vehicle as the origin; For each data point in the spatiotemporal three-dimensional grid, the vehicle occupancy probability of the data point is obtained based on the set of assumed trajectories of the vehicle. For each data point in the spatiotemporal three-dimensional grid, the target occupancy probability of the data point is obtained based on the set of assumed trajectories of each dynamic target. For each data point in the spatiotemporal three-dimensional grid, if the probability of a vehicle occupying the data point is greater than a first preset threshold and the probability of a target occupying the data point is greater than a second preset threshold, then the data point is identified as a potential conflict point.

9. The method according to claim 1, characterized in that, The construction of a dynamic risk field based on the multiple potential conflict points includes: An initial dynamic risk field is established with the vehicle as the origin; the initial dynamic risk field is two-dimensional data, and each data point in the initial dynamic risk field corresponds to a position coordinate. Based on each potential conflict point, the severity of the potential conflict point is obtained; Based on the multiple potential conflict points and the severity of each potential conflict point, the risk value of each location coordinate in the initial dynamic risk field is obtained; The dynamic risk field is obtained based on the risk value of each location coordinate.

10. A vehicle prediction and guidance device, characterized in that, include: The data acquisition module is used to acquire vehicle status data, driver explicit intent data, environmental perception data, map context data, and navigation path data. The vehicle trajectory prediction module is used to determine a set of hypothetical trajectories for the vehicle based on the vehicle status data, the driver's explicit intent data, the map context data, and the navigation path data. The target trajectory prediction module is used to determine a set of hypothetical trajectories for each dynamic target in the environmental perception data. The first determining module is used to determine multiple potential conflict points based on the hypothetical trajectory set of the vehicle and the hypothetical trajectory set of each dynamic target. A construction module is used to construct a dynamic risk field based on the multiple potential conflict points; The second determining module is used to determine a safe path based on the dynamic risk field; The third determining module is used to determine the guidance instruction based on the predicted collision time and severity of the multiple potential conflict points; The rendering module is used to render the safety path, the dynamic risk field, and the guidance instructions onto the vehicle's A-pillar display screen.

Images