Obstacle cutting-in intention prediction method and vehicle

Abstract

The invention discloses an obstacle cut-in intention prediction method and a vehicle, and the method comprises the steps: judging whether there is an obstacle meeting the initial condition of pre-cut-in in the current environment or not during the operation of the vehicle, and obtaining the historical track of the obstacle if there is the obstacle; according to the historical track, judging whether the obstacle has a lane line attaching trend or not; if yes, the transverse speed weight and the transverse position weight of the obstacle are obtained; and predicting a pre-cut-in intention of the obstacle according to the transverse position weight and the transverse speed weight. According to the method, the slow pre-cut-in intention of the obstacles around the vehicle can be accurately predicted.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a method and vehicle for predicting the intention to cut into obstacles. Background Technology

[0002] In autonomous driving technology, trajectory prediction is a core component of the decision-making and planning module, and its accuracy directly impacts driving safety. On structured roads, a particularly challenging scenario is when a vehicle is driving normally in its current lane, and a vehicle in the adjacent lane (hereinafter referred to as the "target vehicle") slowly and continuously cuts into the vehicle's lane. The intent behind this action is difficult to determine; the target vehicle might only briefly cross the line or make a complete lane change. Existing solutions to this problem can be broadly categorized into two types: rule-based and model-based. Rule-based solutions first determine the vehicle's entry and then generate a predicted trajectory based on its intent. The downstream system then projects the predicted trajectory into a displacement-time graph (ST graph) for deceleration. This approach struggles to address the cut-in boundary issue; a conservative approach leads to delayed braking, while an aggressive approach results in excessive braking, affecting the driver's experience. Model-based solutions rely heavily on training data and the target vehicle's historical trajectory, exhibiting poor performance consistency. Even with the same model version, slight changes in the target vehicle's historical state can lead to significant differences in the predicted trajectory, making optimization for individual situations difficult.

[0003] For vehicles slowly cutting into the adjacent lane, it is difficult to accurately determine their target intention. If the cutting trajectory is given too early, the downstream planning module will prematurely project the target into the displacement-time map, causing false braking and affecting the driving experience. If it is given too late, it will cause late braking or even failure to stop, affecting driving safety. How to balance the driving experience and safety issues has become a difficult problem in the field of autonomous driving. Summary of the Invention

[0004] This application provides a method and vehicle for predicting the intention to cut into obstacles. It can accurately predict the slow pre-cutting intention of obstacles around the vehicle before the cutting trajectory of the obstacle is generated. Then, it can perform deceleration based on the prediction result of the pre-cutting intention, which better balances the issues of passenger comfort and driving safety.

[0005] In a first aspect, the present invention provides the following technical solution through an embodiment of the present invention: A method for predicting the intention to cut into an obstacle includes: during vehicle operation, determining whether there is a target obstacle around the vehicle that meets the pre-cut-in criteria; if so, acquiring historical driving data of the target obstacle collected by the vehicle, the historical driving data including driving data at multiple consecutive sampling time points; based on the historical driving data, determining whether the target obstacle has a tendency to hug the lane line, the tendency to hug the lane line refers to the tendency to move closer to the lane line on the same side; if a tendency to hug the lane line exists, acquiring the state information of the target obstacle, and determining the lateral velocity weight and lateral position weight of the target obstacle based on the state information of the target obstacle; and determining the pre-cut-in intention of the target obstacle based on the lateral position weight and the lateral velocity weight.

[0006] Preferably, the driving data at each sampling time point includes: the lateral position of the front corner of the target obstacle and the heading angle of the target obstacle. The step of determining whether the target obstacle exhibits a line-hugging trend based on the historical driving data includes: extracting multiple sets of sub-data from the historical driving data based on the sampling time length of the historical driving data. Each set of sub-data includes driving data for a sampling time period, and the sampling time points of adjacent sets of sub-data are continuous, have a preset time interval, or overlap; for each set of sub-data, based on the sub-data, obtaining the average lateral position of all sampling points included in the sub-data, and the average heading angle of all sampling points within each set of sub-data; and determining whether the target obstacle exhibits a line-hugging trend based on the average lateral position and average heading angle of each set of sub-data.

[0007] Preferably, the step of extracting multiple sets of sub-data from the historical driving data based on the sampling time length of the historical driving data includes: determining the size of the corresponding sliding window and the sliding step size based on the sampling time length; and traversing the historical driving data based on the sliding window and the sliding step size to extract multiple sets of sub-data from the historical driving data.

[0008] Preferably, determining whether the target obstacle has a line-hugging trend based on the average lateral position and average heading angle of each group of sub-data includes: determining whether the difference between the average lateral position of the j-th group of sub-data and the average lateral position of the (j-1)-th group of sub-data is greater than a lateral position threshold, and determining whether the average heading angle of the j-th group of sub-data is less than a heading angle threshold, where j is greater than or equal to 2; if both conditions are met, the single-step approach count is incremented by 1; if any condition is not met, the single-step approach count remains unchanged, and the historical driving data is traversed to obtain the accumulated single-step approach count; determining whether the single-step approach count is greater than a counting threshold, and if it is greater than the counting threshold, determining that the target obstacle has a line-hugging trend.

[0009] Preferably, the step of traversing the historical driving data based on the sliding window and the sliding step size to extract multiple sets of sub-data from the historical driving data includes: taking the current time as the starting time, traversing the historical driving data according to the time order from back to front, based on the sliding window and the sliding step size, and extracting multiple sets of sub-data from the historical driving data.

[0010] Preferably, determining the size and sliding step of the corresponding sliding window based on the sampling time length includes: if the sampling time length is greater than a set length, then extracting a segment of the set length from the sampling time length as a target time length, and determining the sliding window as a first sliding window and the sliding step as a first sliding step based on the target time length; if the sampling time length is less than or equal to the set length, then taking the sampling time length as the target time length, and determining the sliding window as a second sliding window and the sliding step as a second sliding step based on the target time length, wherein the second sliding window is less than or equal to the first sliding window, and the second sliding step is less than or equal to the first sliding step.

[0011] Preferably, the state information includes the longitudinal velocity of the target obstacle, the lateral velocity of the target obstacle, and the type of the target obstacle. Determining the lateral velocity weight of the target obstacle based on its state information includes: determining an upper limit and a lower limit for the lateral velocity of the target obstacle based on its longitudinal velocity; determining a type weight for the target obstacle based on its type; and determining the lateral velocity weight of the target obstacle based on the upper limit, the lower limit, the lateral velocity, a single-step approach count, a set single-step approach count weight, and the type weight.

[0012] Preferably, determining the lateral speed weight of the target obstacle based on the upper limit of lateral speed, the lower limit of lateral speed, the lateral speed, the single-step approach count, the set single-step approach count weight, and the type weight includes: normalizing the lateral speed from the front corner of the target obstacle to the lane line on the same side based on the upper limit of lateral speed, the lower limit of lateral speed, and the lateral speed; and multiplying the normalized result, the single-step approach count weight, the single-step approach count, and the type weight to obtain the lateral speed weight of the target obstacle.

[0013] Preferably, determining the upper limit and lower limit of the lateral velocity of the target obstacle based on its longitudinal velocity includes: multiplying the longitudinal velocity by a set first velocity coefficient to determine the upper limit of the lateral velocity corresponding to the target obstacle; and multiplying the longitudinal velocity by a set second velocity coefficient to determine the lower limit of the lateral velocity corresponding to the target obstacle, wherein the second velocity coefficient is less than the first velocity coefficient.

[0014] Preferably, the status information includes the width of the target obstacle, the width of the lane where the target obstacle is located, the longitudinal distance of the target obstacle relative to the vehicle, the type of the target obstacle, and the lateral position of the front corner of the target obstacle. Determining the lateral position weight of the target obstacle based on its status information includes: dividing the width of the target obstacle by the lane width to obtain the proportion of the target obstacle on the lane; determining an upper and lower limit value for the lateral position of the target obstacle based on the proportion, the longitudinal distance of the target obstacle relative to the vehicle, and the type of the target obstacle; and determining the lateral position weight of the target obstacle based on the upper and lower limit values, the lateral position, the set single-step approach count weight, and the single-step approach count.

[0015] Preferably, determining the lateral position weight of the target obstacle based on the upper limit of the lateral position, the lower limit of the lateral position, the lateral position, the set single-step approach count weight, and the single-step approach count includes: normalizing the lateral distance from the front corner of the target obstacle to the lane line on the same side based on the upper limit of the lateral position, the lower limit of the lateral position, and the lateral position; and multiplying the normalized result, the set single-step approach count weight, and the single-step approach count to obtain the lateral position weight of the target obstacle.

[0016] Preferably, predicting the pre-entry intention of the target obstacle based on the lateral position weight and the lateral velocity weight includes: if the product of the lateral velocity weight and the lateral position weight is greater than a first threshold, or the lateral velocity weight is greater than a second threshold and the lateral position weight is greater than a third threshold, or the lateral velocity weight is equal to 1 and the lateral position weight is greater than a fourth threshold, or the lateral position weight is equal to 1 and the lateral velocity weight is greater than a fifth threshold, then it is determined that the target obstacle has a pre-entry intention.

[0017] Preferably, after determining the pre-cutting intention of the target obstacle, the method further includes: if the target obstacle has a pre-cutting intention, controlling the vehicle to pre-decelerate and determining whether the target obstacle has a cutting intention; if there is a cutting intention, controlling the vehicle to decelerate again.

[0018] Preferably, determining whether there are target obstacles around the vehicle that meet the pre-cut-in access conditions includes: determining whether there are target obstacles in front of and to the side of the vehicle that meet the preset distance conditions; if there are, determining whether the target obstacle is in a lane-changing state and whether the target obstacle has the intention to cut in; if any condition is met, determining that the target obstacle does not meet the pre-cut-in access conditions; if neither condition is met, determining that the target obstacle meets the pre-cut-in access conditions.

[0019] Preferably, before determining whether the target obstacle has a tendency to hug the line based on the historical driving data, the method further includes: determining the sampling time length of the historical driving data, determining whether the sampling time length is greater than or equal to a set minimum length, if yes, then performing the step of determining whether the target obstacle has a tendency to hug the line based on the historical driving data, if no, then ending the prediction of the pre-cutting intention of the target obstacle.

[0020] Secondly, through an embodiment of the present invention, the present invention provides the following technical solution: A vehicle includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the steps of the method described in any of the first aspects above.

[0021] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: The obstacle cutting intention prediction method provided in this invention first determines whether there are obstacles around the vehicle that meet the pre-cutting entry conditions during vehicle operation. If so, it acquires the historical driving data of the obstacle collected by the vehicle. Based on the historical driving data, it determines whether the obstacle has a lane line-hugging trend. If so, it acquires the obstacle's state information and determines the obstacle's lateral velocity weight and lateral position weight based on the state information. Based on the lateral position weight and lateral velocity weight, it determines the obstacle's pre-cutting intention. This method can predict the slow pre-cutting intention of obstacles around the vehicle in advance based on the obstacle's historical driving data and state information before the obstacle's cutting trajectory is given, thereby controlling the vehicle to perform pre-deceleration processing. This effectively avoids the problems caused by traditional technology that relies solely on the cutting trajectory for judgment, such as premature braking due to an early cutting trajectory, which reduces the riding experience, or delayed braking due to a late cutting trajectory, which cannot guarantee driving safety. It also avoids driving integrity problems caused by misjudgment of the cutting intention. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart of the intention prediction method for cutting into obstacles in an embodiment of the present invention; Figure 2 This is a schematic diagram of a scenario involving a vehicle and a target obstacle in an embodiment of the present invention; Figure 3 This is a schematic diagram of the process for determining the trend of the contact line in an embodiment of the present invention; Figure 4 This is a simplified flowchart illustrating the pre-interception intent determination in an embodiment of the present invention. Figure 5 This is a schematic diagram of the vehicle structure in an embodiment of the present invention. Detailed Implementation

[0024] This application provides a method and vehicle for predicting the intention to cut into obstacles. It can accurately predict the slow pre-cutting intention of obstacles around the vehicle before the cutting trajectory of the obstacle is generated. Then, it can perform deceleration based on the prediction result of the pre-cutting intention, which better balances the issues of passenger comfort and driving safety.

[0025] The technical solution of this application embodiment is to solve the above-mentioned technical problems, and the general idea is as follows: A method for predicting the intention to cut into an obstacle includes: during vehicle operation, determining whether there is a target obstacle around the vehicle that meets the pre-cut-in criteria; if so, acquiring historical driving data of the target obstacle collected by the vehicle, the historical driving data including driving data at multiple consecutive sampling time points; based on the historical driving data, determining whether the target obstacle has a tendency to hug the lane line, the tendency to hug the lane line refers to the tendency to move closer to the lane line on the same side; if a tendency to hug the lane line exists, acquiring the state information of the target obstacle, and determining the lateral velocity weight and lateral position weight of the target obstacle based on the state information of the target obstacle; and determining the pre-cut-in intention of the target obstacle based on the lateral position weight and the lateral velocity weight.

[0026] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0027] Firstly, the present invention provides an intention prediction method for cutting into obstacles, specifically, as follows: Figure 1 As shown, the method includes the following steps S101 to S104: Step S101: During vehicle operation, determine whether there are target obstacles around the vehicle that meet the pre-entry access conditions. If there are, acquire the historical driving data of the target obstacle collected by the vehicle. The historical driving data includes driving data at multiple consecutive sampling time points.

[0028] In practical application scenarios, there may be multiple moving objects (hereinafter referred to as obstacles) around the vehicle. Specifically, determining whether there are target obstacles around the vehicle that meet the pre-cut-in access conditions can include: determining whether there are target obstacles in front of and to the side of the vehicle that meet the preset distance conditions; if so, determining whether the target obstacle is in a lane-changing state and whether the target obstacle has the intention to cut in. If any condition is met, it is determined that the target obstacle does not meet the pre-cut-in access conditions; if neither condition is met, it is determined that the target obstacle meets the pre-cut-in access conditions.

[0029] like Figure 2 As shown, E represents the vehicle, S1 represents the target obstacle, min-lon-distance represents the longitudinal distance between the vehicle and the target obstacle (the distance between the rear of the target obstacle and the front of the vehicle), and the lane line between the vehicle and the target obstacle is the same-side lane line Y.

[0030] In this context, "the target obstacle is in a lane-changing state" can refer to situations such as: the turn signal corresponding to the target obstacle is illuminated, or the target obstacle has partially or fully entered the lane currently occupied by the vehicle. "The target obstacle's intention to cut in" can be understood as the target obstacle undergoing a substantial lane departure. "The target obstacle's pre-cutting intention" can be understood as the target obstacle showing a tendency to cut in, but without actually undergoing a lane departure.

[0031] As an example, the vehicle (i.e., the vehicle itself) is equipped with various detection devices such as cameras, radar, and distance sensors to detect in real time the historical driving data (historical trajectory) and status information (heading angle, lateral position, etc. of the target obstacle) of the target obstacle around the vehicle.

[0032] It should be noted that there are multiple ways to obtain the heading angle of the target obstacle. For example, it can be obtained by trajectory fitting, or multiple reference heading angles of the target vehicle can be obtained first. Multiple reference heading angles include at least two of the following: heading angles obtained by principal component analysis algorithm, heading angles obtained by first trajectory fitting, heading angles obtained by second trajectory fitting, and heading angles obtained by trajectory filtering results. Based on the Doppler and azimuth angles of each point in the point cloud of the target vehicle, the variance of the velocity of each point in the point cloud corresponding to each of the multiple reference heading angles is calculated. The reference heading angle corresponding to the smallest variance is determined as the heading angle of the target vehicle. Alternatively, the motion state of the target obstacle can be obtained by the vehicle's onboard sensors (radar, vision, GNSS / IMU), and then the heading angle of the target obstacle obtained based on the motion state can be used.

[0033] In a specific embodiment, the vehicle determines whether there is a moving object in front of and to the side (e.g., in adjacent lanes and adjacent lanes of adjacent lanes) that is within a preset safe distance from the vehicle. If such an object is found, it is identified as a target obstacle. After identifying the target obstacle, the vehicle further determines whether the target obstacle is in a lane-changing state and whether it has an intention to cut in. If either condition is met, i.e., the target obstacle is changing lanes or has an intention to cut in, then no pre-cut-in intention determination is made. If neither condition is met, i.e., the target obstacle is neither in a lane-changing state nor has an intention to cut in, then the pre-cut-in intention determination is performed.

[0034] Optionally, the preset safety distance can be set according to actual needs, for example, 20~40m, and the specific value is not limited in this application.

[0035] Of course, in order to improve the accuracy of detection, after the vehicle determines that there is a moving object that is less than the preset safe distance from the vehicle, it also includes: determining whether the moving object is in the same lane as the vehicle. If so, the moving object is determined to be the target obstacle.

[0036] Specifically, the vehicle can determine whether a moving object is in the same direction as its own by judging whether the object's direction of movement is the same. This allows for targeted detection of moving objects in the same direction, avoiding detection of vehicles in the opposite lane and preventing abnormal detection of oncoming vehicles.

[0037] Specifically, acquiring historical driving data of the target obstacle collected by the vehicle can be achieved by collecting the driving trajectory of the target obstacle once every preset sampling period, resulting in driving data at multiple consecutive sampling time points, such as a preset sampling period of 0.1 to 0.3 seconds.

[0038] Step S102: Based on the historical driving data, determine whether the target obstacle has a tendency to conform to the lane line. The tendency to conform to the lane line refers to the trend of moving closer to the lane line on the same side. Here, the lane line on the same side refers to the lane line on the side of the lane where the target obstacle is located that is closer to the vehicle.

[0039] As an optional embodiment, to ensure that the historical trajectory length meets the minimum requirement and to ensure the accuracy of the prediction, before determining whether the target obstacle has a tendency to hug the line based on the historical driving data, the method may further include: determining the sampling time length of the historical driving data, determining whether the sampling time length is greater than or equal to a set minimum length, if yes, then performing the step of determining whether the target obstacle has a tendency to hug the line based on the historical driving data; if no, then ending the prediction of the target obstacle's pre-entry intention. If the sampling time length is less than the set minimum length, then no pre-entry intention prediction is performed on the target obstacle.

[0040] Optionally, the minimum length can be set between 1 and 3 seconds, such as 2 seconds, and this application does not limit the specific value.

[0041] For example, if the target obstacle has only traveled for 1 second, and the sampling time of the target obstacle's historical travel data at the current moment is less than the set minimum length, then the pre-cut-in intention judgment will not be executed; if the target obstacle has traveled for 10 seconds, then the step of judging whether the target obstacle has a tendency to hug the line will be executed.

[0042] In a specific embodiment, the driving data at each sampling time point includes: the lateral position of the front corner of the target obstacle and the heading angle of the target obstacle. The step of determining whether the target obstacle exhibits a tendency to hug the line based on historical driving data may include: Based on the sampling time length of historical driving data, multiple sets of sub-data are extracted from the historical driving data. Each set of sub-data includes driving data for a sampling time period. The sampling time points of adjacent sets of sub-data are continuous, have a preset time interval, or overlap. For each set of sub-data, the average lateral position of all sampling points included in the sub-data and the average heading angle of all sampling points within each set of sub-data are obtained. Based on the average lateral position and average heading angle of each set of sub-data, it is determined whether the target obstacle has a tendency to hug the lane line. Here, the lateral position refers to the vertical distance between the target obstacle and the lane line on the same side, such as the distance from the left front wheel of the target obstacle to the left lane line, or the distance from the right front wheel of the target obstacle to the right lane line.

[0043] Specifically, heading angle data can be directly acquired through onboard sensors, and then the average heading angle within a sampling time period contained in each set of sub-data can be calculated. For example, a sampling point can be 0.1 seconds, and the sampling time period can be 0.3 seconds. If the sampling time length is 1.5 seconds, and the sampling time points of adjacent sets of sub-data are continuous, 5 sets of sub-data can be extracted from historical driving data. That is, the first set of sub-data is the driving data in the time period of 0.1~0.3 seconds, the second set of sub-data is the driving data in the time period of 0.4~0.6 seconds, the third set of sub-data is the driving data in the time period of 0.7~0.9 seconds, the fourth set of sub-data is the driving data in the time period of 1~1.2 seconds, and the fifth set of sub-data is the driving data in the time period of 1.3~1.5 seconds. Alternatively, the sampling time points of adjacent sub-data groups have a preset duration (such as 0.1 seconds) interval, and four sub-data groups are extracted from historical driving data, namely, the first sub-data group is 0.1~0.3 seconds, the second sub-data group is 0.5~0.7 seconds, the third sub-data group is 0.9~1.1 seconds, and the fourth sub-data group is 1.3~1.5 seconds.

[0044] The average lateral position of each sub-data set is obtained by averaging the multiple lateral positions corresponding to all sampling points in each sub-data set. Similarly, the average heading angle of each sub-data set is obtained by averaging the multiple heading angles corresponding to all sampling points in each sub-data set. For example, if the first sub-data set contains 3 sampling points with lateral positions of 0.3m, 0.35m, and 0.4m respectively, then the average lateral position of this sub-data set is 0.35m.

[0045] As an example, extracting multiple sets of sub-data from historical driving data based on the sampling time length can include: determining the size of the corresponding sliding window and the sliding step size based on the sampling time length; traversing historical driving data based on the sliding window and the sliding step size to extract multiple sets of sub-data from historical driving data.

[0046] In a specific embodiment, determining the size and sliding step of the corresponding sliding window based on the sampling time length can include: if the sampling time length is greater than a set length, then a segment of the set length within the sampling time length is extracted as the target time length, and the sliding window is determined as the first sliding window with a sliding step of the first sliding step length based on the target time length; if the sampling time length is less than or equal to the set length, then the sampling time length is used as the target time length, and the sliding window is determined as the second sliding window with a sliding step of the second sliding step length, wherein the second sliding window is less than or equal to the first sliding window, and the second sliding step length is less than or equal to the first sliding step length. Optionally, the set length can be 1 to 3 seconds, for example, 1 second.

[0047] Specifically, based on the sampling time length of historical driving data, it is divided into different levels, and a corresponding sliding window size (window_size) and sliding step size (window_step) are matched. If the sampling time length is greater than the set length, the time period of the set length is extracted as the target time length, and the first sliding window and the first sliding step size are matched; if the sampling time length is less than or equal to the set length, the sampling time length is directly used as the target time length, and the second sliding window and the second sliding step size are matched, where the second sliding window ≤ the first sliding window, and the second sliding step size ≤ the first sliding step size.

[0048] By determining whether the sampling time length meets the minimum requirement, if it does, the system is divided into different levels based on the sampling time length. This level will affect the subsequent sliding window size design. For example, Level 1 has 10 sampling points, Level 2 has 20 sampling points, 10 sampling points are collected per second, and 3 sampling points constitute the size of a sliding window, with a step size of 1 sampling point. If the minimum requirement is not met, the pre-cutting intention is not determined.

[0049] In a specific embodiment, traversing historical driving data based on a sliding window and a sliding step size to extract multiple sets of sub-data from the historical driving data may include: taking the current time as the starting time, traversing the historical driving data according to the time order from back to front, based on a sliding window and a sliding step size, and extracting multiple sets of sub-data from the historical driving data.

[0050] Assuming the target obstacle's historical trajectory takes 10 seconds, 20 sampling points are collected from the target obstacle's historical driving data. The vehicle's driving intention is determined by these 20 sampling points. During the vehicle's movement, the latest sampling point is continuously used as the initial sampling point, and the historical driving data is traversed in chronological order from back to front.

[0051] For example, if the target obstacle has traveled for 10 seconds, then from the 10th second to the 0th second, the historical driving data is traversed based on a sliding window and a sliding step size. For example, 0.1 seconds is a sampling point, the sliding window size is 3 sampling points, and the sliding step size is 0.1 to 0.5 seconds. Multiple sets of sub-data are extracted from the historical driving data in sequence. For example, the first set of sub-data contains sampling points in the time period of 10 to 9.8 seconds, the second set of sub-data contains sampling points in the time period of 9.7 to 9.5 seconds, the third set of sub-data contains sampling points in the time period of 9.4 to 9.2 seconds, and so on.

[0052] In a specific embodiment, determining whether a target obstacle has a line-hugging trend based on the average lateral position and average heading angle of each group of sub-data can include: determining whether the difference between the average lateral position of the j-th group of sub-data and the average lateral position of the (j-1)-th group of sub-data is greater than a lateral position threshold, and determining whether the average heading angle of the j-th group of sub-data is less than a heading angle threshold, where j is greater than or equal to 2; if both conditions are met, the single-step approach count is incremented by 1; if any condition is not met, the single-step approach count remains unchanged, and the historical driving data is traversed to obtain the accumulated single-step approach count; determining whether the single-step approach count is greater than a counting threshold, and if it is greater than the counting threshold, then it is determined that the target obstacle has a line-hugging trend.

[0053] The lateral position threshold (threshold_l) and heading angle threshold (threshold_heading) can be set according to actual needs. For example, the lateral position threshold can be 0.25~0.3cm, and the heading angle threshold can be ±1°~±3°, that is, the heading angle of the front corner of the vehicle deviates from the center line of the lane by ±1°~±3°.

[0054] Specifically, the process of determining whether a target obstacle exhibits a tendency to hug the line may include: determining whether the initial index value is less than the sampling time length of the historical driving data (i.e., the length of the historical trajectory); if so, then performing the following steps to traverse the historical trajectory in reverse chronological order: The algorithm determines the average lateral position of all sampling points within the current sliding window and obtains the average heading angle of all sampling points within the current sliding window. The position of the current sliding window corresponds to the index value. It checks whether the difference between the average lateral position of the current sliding window and the average lateral position of the previous sliding window is greater than the lateral position threshold, and whether the average heading angle of the current sliding window is less than the heading angle threshold. If both conditions are met, the single-step approach count is incremented by 1. If either condition is not met, the single-step approach count remains unchanged, and the index value is updated to the sum of the previous index value, the size of the sliding window, and the sliding step size. It checks whether the updated index value is less than the length of the historical trajectory. If so, the step of traversing the historical trajectory is repeated until it is determined that the updated index value is greater than or equal to the length of the historical trajectory. Then, it checks whether the single-step approach count is greater than the counting threshold. If it is greater than the counting threshold, it is determined that the obstacle has a lane line rubbing trend. If it is less than or equal to the counting threshold, it is determined that the obstacle does not have a lane line rubbing trend.

[0055] Specifically, such as Figure 3 As shown, the size of the sliding window (window_size) and the sliding step (window_step) are determined according to the sampling time length. The main loop is then entered to check if the current index value (index) is less than the history length (history_size). If not, the main loop is exited and the next step is performed. If the condition is met, the system continues to check if i is less than the size of the sliding window (window_size, for example, 3 sampling points). If i is greater than or equal to the size of the sliding window, the index value (index) is the size of the sliding window plus the sliding step, resulting in the updated index value (index), and j+1. The main loop is then entered again based on the updated index value (index). If i is less than the size of the sliding window (window_size), the average lateral position (front_corner_l[j]) and the average heading angle (heading[j]) of the j-th sub-data set are calculated, where j is the index value of the sub-data set.

[0056] Starting from the second group of sub-data, determine whether the difference between the average lateral position of the j-th group of sub-data and the average lateral position of the (j-1)-th group of sub-data is greater than the preset lateral position threshold_l, and whether the average heading angle of the j-th group of sub-data is less than the preset heading angle threshold_heading. If both conditions are met, increment the single-step approach count by 1. If either condition is not met, keep the single-step approach count unchanged. Iterate through all sub-data according to the above rules to obtain the cumulative single-step approach count.

[0057] The system checks if the accumulated single-step approach count is greater than a preset threshold `threshold_count`. If it is, the target obstacle is determined to have a tendency to approach the lane (`is_approaching_lane=true`). If it is less than or equal to the threshold, the target obstacle is determined not to have a tendency to approach the lane, and the pre-cutting intent prediction ends. Here, `i` represents the index value of the sliding window.

[0058] By using a sliding window method to traverse the features of historical trajectories, it is determined whether the target obstacle has a tendency to approach the lane line. By calculating the average lateral position and average heading angle within each sliding window, the difference between the average lateral positions of adjacent sliding windows is used to determine whether the target obstacle is close to or far from the lane line on the same side. Then, by combining the average heading angle and heading angle threshold, it is determined whether the obstacle has a tendency to approach the lane line on the same side, thus obtaining whether the target obstacle has a tendency to approach the lane line.

[0059] As another optional embodiment, extracting multiple sets of sub-data from historical driving data based on the sampling time length of historical driving data may include: dividing the sampling time length of historical driving data into equal intervals according to a preset division rule, dividing the sampling time length into multiple equal-length sampling time periods, and extracting multiple sets of sub-data, with each set of sub-data corresponding to a sampling time period.

[0060] Step S103: If there is a trend of adhering to the line, obtain the state information of the target obstacle, and determine the lateral velocity weight and lateral position weight of the target obstacle based on the state information of the target obstacle. Step S104: Determine the pre-cutting intention of the target obstacle based on the lateral position weight and the lateral velocity weight.

[0061] In a specific embodiment, the state information includes the longitudinal velocity of the target obstacle, the lateral velocity of the target obstacle, and the type of the target obstacle. Determining the lateral velocity weight of the target obstacle based on its state information may include: Based on the longitudinal velocity of the target obstacle, determine the upper and lower limits of its lateral velocity; based on the type of the target obstacle, determine its type weight; based on the upper and lower limits of its lateral velocity, its lateral velocity, single-step approach count, the set single-step approach count weight, and its type weight, determine the lateral velocity weight of the target obstacle.

[0062] As an example, determining the type weight of the target obstacle based on its type obj_type can include: if the target obstacle is of type 1, then the type weight of the target obstacle is determined to be the first weight; if the target obstacle is of type 2, then the type weight of the target obstacle is determined to be the second weight; and if the target obstacle is of type 3, then the type weight of the target obstacle is determined to be the third weight.

[0063] Optionally, the first type can be large trucks, tanker trucks, etc.; the second type can be cars, off-road vehicles, SUVs, etc.; and the third type can be Vulnerable Road Users (VRUs), such as electric vehicles, motorcycles, bicycles, etc. The weight of the first type is greater than or equal to the weight of the third type, and the weight of the second type is less than the weight of the third type.

[0064] Determining the upper and lower limits of the lateral velocity of the target obstacle based on its longitudinal velocity may include: determining the upper limit of the lateral velocity of the target obstacle by multiplying the longitudinal velocity by a set first velocity coefficient; and determining the lower limit of the lateral velocity of the target obstacle by multiplying the longitudinal velocity by a set second velocity coefficient, wherein the second velocity coefficient is less than the first velocity coefficient.

[0065] As an example, determining the upper and lower limits of lateral velocity includes: low_lat_velocity=k1× _speed high_lat_velocity=k2× _speed Where k1 is the first velocity coefficient, k2 is the second velocity coefficient, low_lat_velocity is the lower limit of lateral velocity, and high_lat_velocity is the upper limit of lateral velocity. _speed represents the longitudinal speed. k1 and k2 can be set values.

[0066] In a specific embodiment, the lateral speed weight of the target obstacle is determined based on the upper limit of lateral speed, the lower limit of lateral speed, lateral speed, single-step approach count, a set single-step approach count weight, and a type weight. This can include: normalizing the lateral speed from the front corner of the target obstacle to the lane line on the same side based on the upper limit of lateral speed, the lower limit of lateral speed, and lateral speed; and multiplying the normalized result, the single-step approach count weight, the single-step approach count, and the type weight to obtain the lateral speed weight of the target obstacle.

[0067] First, determine the upper and lower limits of the lateral velocity based on the longitudinal velocity range of the obstacle. Then, normalize the lateral velocity based on the upper and lower limits. Finally, calculate the lateral velocity weight by combining the single-step approach count.

[0068] Specifically, the calculation process for the lateral velocity weight includes: lat_velocity_probability= ×count_weight_lat×count×obj_type_weight Where lat_velocity is the lateral velocity, count_weight_lat is the single-step proximity count weight (a set value), obj_type_weight is the target type weight, high_lat_velocity is the upper limit of the lateral velocity, low_lat_velocity is the lower limit of the lateral velocity, lat_velocity_probability is the lateral velocity weight, and count is the single-step proximity count. The lateral velocity is mapped to the range (1.0-0.3).

[0069] It should be noted that the lateral velocity is the lateral velocity of the front corner of the target obstacle relative to the lane line on the same side (the rate of change of velocity along the vertical direction of the vehicle); the longitudinal velocity is the longitudinal velocity of the front corner of the target obstacle relative to the lane line on the same side (the rate of change of velocity along the direction of vehicle travel).

[0070] In a specific embodiment, the state information includes the width of the target obstacle, the width of the lane where the target obstacle is located, the longitudinal distance of the target obstacle relative to the vehicle, the type of the target obstacle, and the lateral position of the front corner of the target obstacle. Based on the current driving data, the lateral position weight of the target obstacle is obtained, which may include: dividing the width of the target obstacle by the width of the lane to obtain the proportion of the target obstacle on the lane; determining the upper limit and lower limit of the lateral position of the target obstacle based on the proportion, the longitudinal distance of the target obstacle relative to the vehicle, and the type of the target obstacle; and determining the lateral position weight of the target obstacle based on the upper limit, lower limit, lateral position, set single-step approach count weight, and single-step approach count.

[0071] Specifically, the width of the target obstacle and the lane width can be detected by cameras or position detectors installed on the vehicle, and the proportion of the target obstacle in the lane can be calculated using the following formula: obj_lane_width_ratio=obj_width÷cur_lane_width Where obj_lane_width_ratio is the lane ratio of the target obstacle, obj_width is the width of the target obstacle, and cur_lane_width is the width of the lane where the target obstacle is located.

[0072] Next, based on the proportion, the longitudinal distance lon_distance of the target obstacle relative to the vehicle, and the type of the target obstacle, the upper limit value of the lateral position low_front_corner_l and the lower limit value of the lateral position high_front_corner_l of the target obstacle are determined, which may include: Based on the lane ratio, the upper and lower limits of the first lateral position of the obstacle are determined, wherein the upper and lower limits of the first lateral position are inversely proportional to the lane ratio. Based on the longitudinal distance of the obstacle relative to the vehicle, the upper and lower limits of the second lateral position of the obstacle are determined, wherein the upper and lower limits of the second lateral position are inversely proportional to the longitudinal distance. Based on the type of obstacle, the upper and lower limits of the third lateral position of the obstacle are determined. Based on the upper, lower, and third lateral position limits, the upper limit of the lateral position of the obstacle is determined, and based on the lower, lower, and third lateral position limits, the lower limit of the lateral position of the obstacle is determined.

[0073] Specifically, the calculation of the upper and lower limits of the lateral position is related to three factors: the longitudinal distance (lon_distance) of the target obstacle relative to the vehicle, the lane width ratio (obj_lane_width_ratio) of the obstacle, and the obstacle type (obj_type). A larger longitudinal distance (lon_distance) results in smaller lower and upper limits of the lateral position (low_front_corner_l and high_front_corner_l); a smaller lane width ratio (obj_lane_width_ratio) results in larger lower and upper limits of the lateral position (low_front_corner_l and high_front_corner_l); and larger lower and upper limits of the lateral position (low_front_corner_l) for larger vehicles and VRUs compared to smaller vehicles.

[0074] In a specific embodiment, the lateral position weight of the target obstacle is determined based on the upper limit of the lateral position, the lower limit of the lateral position, the lateral position, the set single-step approach count weight, and the single-step approach count. This may include: normalizing the lateral distance from the front corner of the target obstacle to the lane line on the same side based on the upper limit of the lateral position, the lower limit of the lateral position, and the lateral position; and multiplying the normalized result, the set single-step approach count weight, and the single-step approach count to obtain the lateral position weight of the target obstacle.

[0075] Specifically, first determine the proportion of the target obstacle in the lane. Then, based on the longitudinal distance of the target obstacle relative to the vehicle, the type of the target obstacle, and its proportion, determine the upper and lower limits of its lateral position. Next, normalize the lateral distance from the front corner of the target obstacle to the lane line on the same side based on these limits. Finally, calculate the lateral position weight using a single-step approach count. The lateral position weight can be calculated using the following formula: front_corner_l_probability= ×count_weight_lat×count Where `low_front_corner_l` is the lower limit of the horizontal position, `high_front_corner_l` is the upper limit of the horizontal position, `front_corner_l` is the horizontal position of the front corner, `front_corner_l_probability` is the horizontal position weight, `count_weight_lat` is the single-step proximity count weight, and `count` is the single-step proximity count. The horizontal position is mapped to the interval (1.0-0.3).

[0076] In a specific embodiment, predicting the pre-entry intention of the target obstacle based on the lateral position weight and the lateral velocity weight may include: If the product of the lateral velocity weight and the lateral position weight is greater than the first threshold, or the lateral velocity weight is greater than the second threshold and the lateral position weight is greater than the third threshold, or the lateral velocity weight is equal to 1 and the lateral position weight is greater than the fourth threshold, or the lateral position weight is equal to 1 and the lateral velocity weight is greater than the fifth threshold, then it is determined that the target obstacle has a pre-cutting intention. The fourth threshold is less than the third threshold, and the fifth threshold is less than the second threshold.

[0077] Specifically, after determining that the target obstacle has a tendency to hug the line, the next step is to determine whether the lateral velocity weight and lateral position weight of the target obstacle both meet the set requirements. The determination of whether the lateral velocity weight and lateral position weight meet the requirements is based on the following four conditions being satisfied: Lateral velocity weight lat_velocity_probability × lateral position weight front_corner_l_probability > first threshold 1; Alternatively, the lateral velocity weight lat_velocity_probability > the second threshold threshold2, and the lateral position weight front_corner_l_probability > the third threshold threshold3; Alternatively, the lateral velocity weight lat_velocity_probability = 1, and the lateral position weight front_corner_l_probability > the fourth threshold 4; Alternatively, the lateral position weight front_corner_l_probability=1, and the lateral velocity weight lat_velocity_probability>the fifth threshold5; Among them, the first threshold (threshold1), the second threshold (threshold2), the third threshold (threshold3), the fourth threshold (threshold4), and the fifth threshold (threshold5) are the thresholds for the corresponding weights, and all are calibration values.

[0078] If all conditions are met, then pre-cutin is valid.

[0079] This method achieves both lateral verification: determining whether an obstacle cuts in from a distant lateral position; if so, no pre-cutin judgment is made; and longitudinal verification: determining the longitudinal action distance based on the type of the target obstacle and its lateral velocity, where the larger the vehicle type, the smaller the lateral velocity.

[0080] like Figure 4 As shown, the process mainly includes: the vehicle performs scene recognition, determines whether there is a target obstacle in the current scene that meets the pre-cutin admission conditions, if there is, it continues to determine whether the target obstacle has a tendency to hug the line, if there is a tendency to hug the line, it calculates the lateral position weight and lateral velocity weight, and determines the pre-cutin intention of the target obstacle based on the lateral position weight and lateral velocity weight.

[0081] The pre-entry intention of the target obstacle includes both no entry intention and entry intention. Specifically, after determining the pre-entry intention of the target obstacle, the process may further include: if the target obstacle has a pre-entry intention, controlling the vehicle to pre-decelerate and determining whether the target obstacle has an entry intention; if there is an entry intention, controlling the vehicle to decelerate again.

[0082] For example, pre-deceleration here could be deceleration when the brake pedal is in a first opening range, and further deceleration could be deceleration when the brake pedal is in a second opening range. The maximum value in the first opening range is less than or equal to the minimum value in the second opening range. This allows for a slight pre-deceleration to prevent the vehicle from cutting into the obstacle when its intention to do so is anticipated, followed by a deeper deceleration once the intention is confirmed. This gradual deceleration improves both driving safety and passenger comfort. For instance, the first opening range could be 10% to 60%, such as 20%, 30%, or 40% pedal opening, and the second opening range could be 60% to 90%, such as 60%, 70%, or 80% pedal opening.

[0083] By adding the judgment of pre-cutting intention, the entry of the target obstacle can be more accurately identified in advance. The method of combining pre-cutting and cut-in intention judgment is adopted. The cut-in intention is used to guide the generation of the cut-in trajectory, and the pre-cutting is used to provide the downstream module before the cut-in trajectory is given, so that the downstream can pre-decelerate in advance. This can appropriately relax the cut-in intention judgment conditions, reduce the false braking caused by the early cutting trajectory, and also reduce the occurrence of false cut-in. At the same time, it can also take into account the problem of untimely braking caused by the late cutting trajectory.

[0084] This avoids the problem of judging entry intentions based on a wide range when pre-entry intention judgment is not set. For example, if the distance between the target obstacle and the lane line on the same side is less than or equal to 0.3m, and the heading angle of the target obstacle meets certain conditions, the target obstacle is directly judged to meet the entry intention. In this case, because the range of entry conditions is set too wide, there may be problems such as misjudgment of entry or premature braking, which will affect the passenger experience. With the addition of pre-entry intention judgment, entry intention can be judged based on more precise conditions. For example, if the distance between the target obstacle and the lane line on the same side is less than or equal to 0.2m, and the heading angle meets certain conditions, the target obstacle is judged to meet the entry intention.

[0085] Therefore, this application proposes a dual-intention judgment scheme of pre-cut-in intention and cut-in intention. The cut-in intention is used to guide the generation of the cut-in trajectory, while the pre-cut-in intention is given to the downstream module before the cut-in trajectory is given, so that the downstream module can perform pre-deceleration in advance. This can appropriately relax the cut-in intention judgment conditions, reduce the false braking caused by the early cutting-in trajectory, reduce the occurrence of false cut-in, and at the same time take into account the problem of untimely braking caused by the late cutting-in trajectory.

[0086] In summary, the obstacle cutting intention prediction method provided by the embodiments of the present invention can accurately predict the slow pre-cutting intention of obstacles around the vehicle before the obstacle cutting trajectory is generated. Then, based on the pre-cutting intention, deceleration can be performed at an appropriate time, which better balances the riding experience and driving safety.

[0087] Secondly, based on the same inventive concept, this embodiment provides a vehicle 500, such as... Figure 5 As shown, it includes: a memory 501, a processor 502, and a computer program 503 stored in the memory and executable on the processor. When the processor 502 executes the program, it implements the steps of the intention prediction method described in the first aspect above.

[0088] Since the vehicle described in this embodiment is the vehicle used to implement the intent prediction method in the embodiments of this application, those skilled in the art can understand the specific implementation method and various variations of the vehicle in this embodiment based on the intent prediction method described in the embodiments of this application. Therefore, how the vehicle implements the method in the embodiments of this application will not be described in detail here. Any vehicle used by those skilled in the art to implement the intent prediction method in the embodiments of this application falls within the scope of protection of this application.

[0089] 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, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0090] 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 1 One or more processes and / or boxes Figure 1 A module that specifies the function in one or more boxes.

[0091] 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 modules implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0092] 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 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0093] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0094] 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 method for predicting the intent to cut into an obstacle, characterized in that, include: During vehicle operation, it is determined whether there are target obstacles around the vehicle that meet the pre-entry access conditions. If there are, the historical driving data of the target obstacle collected by the vehicle is obtained. The historical driving data includes driving data at multiple consecutive sampling time points. Based on the historical driving data, it is determined whether the target obstacle has a tendency to hug the lane line, where the tendency to hug the lane line refers to the tendency to move closer to the lane line on the same side. If a line-hugging trend exists, the state information of the target obstacle is obtained, and the lateral velocity weight and lateral position weight of the target obstacle are determined based on the state information of the target obstacle. The pre-entry intention of the target obstacle is determined based on the lateral position weight and the lateral velocity weight.

2. The method as described in claim 1, characterized in that, The driving data at each sampling time point includes: the lateral position of the front corner of the target obstacle and the heading angle of the target obstacle. The step of determining whether the target obstacle exhibits a tendency to hug the line based on the historical driving data includes: Based on the sampling time length of the historical driving data, multiple sets of sub-data are extracted from the historical driving data. Each set of sub-data includes driving data for a sampling time period. The sampling time points of adjacent sets of sub-data are continuous, have a preset time interval, or overlap. For each set of sub-data, based on the sub-data, obtain the average lateral position of all sampling points included in the sub-data, and the average heading angle of all sampling points in each set of sub-data; Based on the average lateral position and average heading angle of each set of sub-data, it is determined whether the target obstacle has a tendency to hug the line.

3. The method as described in claim 2, characterized in that, The step of extracting multiple sets of sub-data from the historical driving data based on the sampling time length of the historical driving data includes: The size of the corresponding sliding window and the sliding step size are determined based on the sampling time length. Based on the sliding window and the sliding step size, the historical driving data is traversed to extract multiple sets of sub-data from the historical driving data.

4. The method as described in claim 2, characterized in that, The step of determining whether the target obstacle has a tendency to hug the line based on the average lateral position and average heading angle of each set of sub-data includes: Determine whether the difference between the average lateral position of the j-th sub-data group and the average lateral position of the (j-1)-th sub-data group is greater than the lateral position threshold, and determine whether the average heading angle of the j-th sub-data group is less than the heading angle threshold, where j is greater than or equal to 2; If all conditions are met, the single-step approach count is incremented by 1. If any condition is not met, the single-step approach count remains unchanged. The historical driving data is traversed to obtain the accumulated single-step approach count. Determine whether the single-step approach count is greater than the counting threshold. If it is greater than the counting threshold, then determine that the target obstacle has a tendency to adhere to the line.

5. The method as described in claim 3, characterized in that, The process involves traversing the historical driving data based on the sliding window and the sliding step size, and extracting multiple sets of sub-data from the historical driving data, including: Starting from the current time, the historical driving data is traversed in reverse chronological order based on the sliding window and the sliding step size, and multiple sets of sub-data are extracted from the historical driving data.

6. The method as described in claim 3, characterized in that, The step of determining the size and step size of the corresponding sliding window based on the sampling time length includes: If the sampling time length is greater than the set length, then a period of the set length is extracted from the sampling time length as the target time length, and the sliding window is determined as the first sliding window and the sliding step size is the first sliding step size based on the target time length; If the sampling time length is less than or equal to the set length, then the sampling time length is taken as the target time length, and the sliding window is determined as the second sliding window and the sliding step size is determined as the second sliding step size according to the target time length, wherein the second sliding window is less than or equal to the first sliding window and the second sliding step size is less than or equal to the first sliding step size.

7. The method according to any one of claims 1 to 5, characterized in that, The state information includes the longitudinal velocity of the target obstacle, the lateral velocity of the target obstacle, and the type of the target obstacle. Determining the lateral velocity weight of the target obstacle based on its state information includes: Based on the longitudinal velocity of the target obstacle, the upper limit and lower limit of the lateral velocity of the target obstacle are determined; Based on the type of the target obstacle, the type weight of the target obstacle is determined; The lateral speed weight of the target obstacle is determined based on the upper limit of lateral speed, the lower limit of lateral speed, the lateral speed, the single-step approach count, the set single-step approach count weight, and the type weight.

8. The method as described in claim 7, characterized in that, The determination of the lateral velocity weight of the target obstacle based on the upper limit of lateral velocity, the lower limit of lateral velocity, the lateral velocity, the single-step approach count, the set single-step approach count weight, and the type weight includes: Based on the upper limit of lateral velocity, the lower limit of lateral velocity, and the lateral velocity, the lateral velocity from the front corner of the target obstacle to the lane line on the same side is normalized. The lateral velocity weight of the target obstacle is obtained by multiplying the normalized result, the single-step approach count weight, the single-step approach count, and the type weight.

9. The method as described in claim 7, characterized in that, The step of determining the upper limit and lower limit of the lateral velocity of the target obstacle based on its longitudinal velocity includes: The product of the longitudinal velocity and the set first velocity coefficient is determined as the upper limit of the lateral velocity corresponding to the target obstacle; The product of the longitudinal velocity and the set second velocity coefficient is determined as the lower limit of the lateral velocity corresponding to the target obstacle, wherein the second velocity coefficient is less than the first velocity coefficient.

10. The method according to any one of claims 1 to 5, characterized in that, The status information includes the width of the target obstacle, the width of the lane where the target obstacle is located, the longitudinal distance of the target obstacle relative to the vehicle, the type of the target obstacle, and the lateral position of the front corner of the target obstacle. Based on the status information of the target obstacle, the lateral position weight of the target obstacle is determined, including: The width of the target obstacle is divided by the width of the lane to obtain the proportion of the target obstacle on the lane; Based on the percentage, the longitudinal distance of the target obstacle relative to the vehicle, and the type of the target obstacle, the upper limit and lower limit of the lateral position of the target obstacle are determined. The lateral position weight of the target obstacle is determined based on the upper limit of the lateral position, the lower limit of the lateral position, the lateral position, the set single-step approach count weight, and the single-step approach count.

11. The method as described in claim 10, characterized in that, The step of determining the lateral position weight of the target obstacle based on the upper limit of the lateral position, the lower limit of the lateral position, the lateral position, the set single-step approach count weight, and the single-step approach count includes: Based on the upper limit of the lateral position, the lower limit of the lateral position, and the lateral position, the lateral distance from the front corner of the target obstacle to the lane line on the same side is normalized. The lateral position weight of the target obstacle is obtained by multiplying the normalized result, the set single-step approach count weight, and the single-step approach count.

12. The method as described in claim 1, characterized in that, The step of predicting the pre-entry intention of the target obstacle based on the lateral position weight and the lateral velocity weight includes: If the product of the lateral velocity weight and the lateral position weight is greater than a first threshold, or the lateral velocity weight is greater than a second threshold and the lateral position weight is greater than a third threshold, or the lateral velocity weight is equal to 1 and the lateral position weight is greater than a fourth threshold, or the lateral position weight is equal to 1 and the lateral velocity weight is greater than a fifth threshold, then it is determined that the target obstacle has a pre-cutting intention.

13. The method as described in claim 1, characterized in that, After determining the pre-entry intention of the target obstacle, the method further includes: If the target obstacle has a pre-cutting intention, the vehicle is controlled to pre-decelerate, and it is determined whether the target obstacle has a cutting intention. If there is an intention to cut in, the vehicle will slow down again.

14. The method as described in claim 1, characterized in that, The determination of whether there are target obstacles around the vehicle that meet the pre-entry access conditions includes: Determine whether there are target obstacles in front of and to the side of the vehicle that meet the preset distance conditions; If it exists, determine whether the target obstacle is in a lane-changing state and whether the target obstacle has an intention to cut in. If either condition is met, it is determined that the target obstacle does not meet the pre-cut-in access conditions. If neither condition is met, it is determined that the target obstacle meets the pre-cut-in access conditions.

15. The method as described in claim 1, characterized in that, Before determining whether the target obstacle has a tendency to hug the line based on the historical driving data, the method further includes: Determine the sampling time length of the historical driving data, and determine whether the sampling time length is greater than or equal to the set minimum length. If so, execute the step of determining whether the target obstacle has a tendency to hug the line based on the historical driving data. If not, end the prediction of the pre-cutting intention of the target obstacle.

16. A vehicle, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, performs the steps of the method according to any one of claims 1 to 15.

Images