A real-time collision detection method based on driving intention and trajectory prediction

Abstract

The application discloses a real-time collision detection method based on driving intention and trajectory prediction, constructs an intention-based trajectory prediction framework without deep learning, constructs a driving intention prediction model based on features extracted from non-intersection scenes and intersection scenes respectively, and adopts a Bayesian optimization algorithm based on a tree-structured Parzen estimator (TPE) to complete training of the driving intention prediction model, adopts the trained driving intention prediction model to predict driving intentions of obstacle vehicles around a controlled vehicle, and then predicts driving trajectories of the obstacle vehicles around the controlled vehicle for the non-intersection scenes and the intersection scenes respectively, so that whether the controlled vehicle and the obstacle vehicles will collide is judged through comparison and calculation of the driving trajectories of the controlled vehicle and the obstacle vehicles, and the calculation real-time performance of the collision detection is effectively improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of unmanned vehicle prediction planning, and particularly relates to a real-time collision detection method based on driving intention and trajectory prediction. BACKGROUND

[0002] Collision detection is an important task in autonomous driving, and accurate and reliable collision detection is crucial to ensure the safety of autonomous vehicles. At present, many deep learning-based algorithms have achieved good performance in the field of trajectory prediction, but some of these algorithms are not suitable for deployment in engineering projects. The reasons are as follows: on the one hand, considering the numerous factors that affect the future trajectory of a vehicle, including surrounding vehicles, pedestrians, lane markings, and complex multi-modal features, due to the excessive number of parameters involved, the forward inference time of the model often cannot meet the real-time requirements. In autonomous driving, even the slightest delay can lead to catastrophic accidents, and the demand for real-time performance limits the applicability of the most advanced methods. On the other hand, most mainstream trajectory prediction algorithms can only predict trajectories 3-5 seconds into the future, which is not sufficient to meet the safety needs of autonomous vehicles in certain high-speed scenarios. In addition, due to the inevitable deviation in the predicted trajectory of surrounding vehicles, especially in the case of long prediction horizon, simply relying on determining whether two vehicles will collide at the same time for collision detection can lead to unreliable collision detection results. SUMMARY

[0003] Therefore, the application provides a real-time collision detection method based on driving intention and trajectory prediction, which realizes real-time detection of collision risk between a controlled vehicle and an obstacle vehicle.

[0004] The real-time collision detection method based on driving intention and trajectory prediction provided by the application specifically includes the following steps:

[0005] Step 1: Obtain vehicle trajectory information and its annotated vehicle driving intention true value to construct a training sample data set;

[0006] Step 2: Establish a driving intention prediction model based on a single-layer LSTM network of a non-deep network, which is used to predict the driving intention of an obstacle vehicle located around the controlled vehicle. The input of the driving intention prediction model is as follows:

[0007]

[0008] wherein, I i is the historical trajectory sequence feature of the ith obstacle vehicle, is the feature of the ith obstacle vehicle at time t; the feature of a non-intersection scene is represented as as follows:

[0009]

[0010] wherein, is the distance along the road reference line from the start of the road to the current vehicle position at time t-k, is the offset along the road reference line perpendicular to the current position at time t-k, is the offset of the obstacle vehicle relative to the current heading angle of the controlled vehicle at time t-k, is the speed value of the obstacle vehicle at time t-k, is the acceleration value of the obstacle vehicle at time t-k;

[0011] The feature representation of the intersection scenario is as shown in the following formula:

[0012]

[0013] wherein, and is the position coordinate of the i-th obstacle vehicle in the plane rectangular coordinate system at time t-k under the given UTM coordinate system, is the speed value of the obstacle vehicle along the X-axis at time t-k, is the speed value of the obstacle vehicle along the Y-axis at time t-k;

[0014] Step 3, the optimal hyperparameters of the driving intention prediction model are obtained by using the Bayesian optimization algorithm based on tree Parzen estimation, and the training of the driving intention prediction model is completed;

[0015] Step 4, in actual use, the vehicle trajectory information of the obstacle vehicle is taken as the input of the driving intention prediction model obtained by training, the driving intention of the obstacle vehicle is calculated by the driving intention prediction model, and the driving trajectory of the obstacle vehicle is predicted for non-intersection scenarios and intersection scenarios respectively, and the trajectory prediction range is a set period;

[0016] Step 5, the oriented bounding box of the controlled vehicle and the obstacle vehicle related thereto is determined, and the length and width of the oriented bounding box are both expanded by k times; then the collision time difference is introduced to represent the time interval of the controlled vehicle and the obstacle vehicle reaching the same position, if the collision time difference of the two vehicles is less than the set threshold δ, it is determined that the two vehicles will collide, otherwise it is determined that the two vehicles will not collide.

[0017] Further, the prediction method of the driving trajectory of the obstacle vehicle for non-intersection scenarios is:

[0018] For vehicles driving in a straight line, it is assumed that the vehicle maintains uniform acceleration motion in the road direction and the speed does not exceed the maximum speed threshold, and continuously approaches the center line of the lane in the direction perpendicular to the road, and the process is represented by the following formula:

[0019]

[0020] where s t is the distance of the obstacle vehicle along the road reference line at time t, s t+1 is the distance of the obstacle vehicle along the road reference line at time t+1, v s is the projection of the speed of the obstacle vehicle in the S direction, a s is the projection of the acceleration of the obstacle vehicle in the S direction, Δt is the prediction time interval, is the L coordinate of the projection point of the obstacle vehicle on the center line at time t, k is the speed of the obstacle vehicle approaching the center line, l t is the offset of the obstacle vehicle along the road reference line perpendicular to the current position at time t, l t+1 is the offset of the obstacle vehicle along the road reference line perpendicular to the current position at time t+1;

[0021] For vehicles with lane changing intention, fourth-order polynomial s(t) and fifth-order polynomial l(t) are used to model the trajectories in the S direction and the L direction, as shown in the following formulas:

[0022]

[0023] s(t) is the distance of the obstacle vehicle along the road reference line, l(t) is the distance of the obstacle vehicle along the direction perpendicular to the road reference line; s(0) = s0 indicates that the first expression for solving s(t) is constructed by substituting s0, s(0) = s0 indicates that the second expression for solving s(t) is constructed by substituting , and so on to solve the coefficients of the s(t) polynomial; l(0) = l0 indicates that the first expression is constructed by substituting l0, s(0) = s0 indicates that the second expression is constructed by substituting , and so on to solve the coefficients of the l(t) polynomial; s0, l0 and their higher-order derivatives are calculated from the current state of the obstacle vehicle; T1 is the lane changing time, and the candidate lane changing trajectory of the obstacle vehicle is calculated by sampling T1 in the Frenet coordinate system; and the optimal trajectory is selected by using the cost function shown in the following formula:

[0024] cost lc (T1) = w1T1 + w2max{a n}

[0025] where w1 and w2 are weight parameters, max{a n} is the maximum normal acceleration in the trajectory, and the trajectory with the minimum cost function value is selected as the lane changing trajectory.

[0026] Further, the manner of predicting the driving trajectory of the obstacle vehicle for the intersection scene is:

[0027] The turning or straight trajectory in the X and Y directions of the UTM coordinate system is modeled respectively, as shown in the following formula:

[0028]

[0029] Wherein, x(0)=x0 represents the state of substituting x0 to construct the first expression of solving x(t), x(0)=x0 represents the state of substituting x0 to construct the first expression of solving x(t), x(0)=x0 represents the state of substituting x0 to construct the first expression of solving x(t), x(0)=x0 represents the state of substituting x0 to construct the first expression of solving x(t), x(0)=x0 represents the state of substituting x0 to construct the first expression of solving x(t), exit x(0)=x0 represents the state of substituting x0 to construct the first expression of solving x(t), exit x(0)=x0 represents the state of substituting x0 to construct the first expression of solving x(t),

[0030] Further, the driving intention prediction model of the non-intersection scene is trained by using the data set NGSIM, and the driving intention prediction model of the intersection scene is trained by using the data set INTERACTION.

[0031] Further, the driving intention includes left turn, right turn or straight driving.

[0032] Further, the set period is 8 seconds.

[0033] Further, the vehicle trajectory information includes vehicle position, speed, acceleration and heading angle.

[0034] Beneficial effects:

[0035] 1、The application constructs an intention-based trajectory prediction framework without deep learning, constructs driving intention prediction models based on the features extracted from non-intersection scenes and intersection scenes respectively, and completes the training of the driving intention prediction models by using a Bayesian optimization algorithm based on a tree-structured Parzen estimator (TPE), predicts the driving intention of the obstacle vehicles around the controlled vehicle by using the trained driving intention prediction model, and respectively predicts the driving trajectories of the obstacle vehicles around the controlled vehicle for non-intersection scenes and intersection scenes, and judges whether the controlled vehicle and the obstacle vehicles will collide by comparing and calculating the driving trajectories of the controlled vehicle and the obstacle vehicles, thereby effectively improving the calculation real-time performance of the collision detection.

[0036] 2、The application introduces time redundancy and space redundancy in the collision detection process, effectively avoiding the problem that the current trajectory-based collision detection algorithm has too strict calculation requirements, and can effectively cope with the calculation challenges that may be faced in real scenarios, and improve the adaptability of the method to the engineering requirements of automatic driving collision detection. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 A driving intention prediction model network structure diagram constructed in a real-time collision detection method based on driving intention and trajectory prediction provided by the application.

[0038] Figure 2 A driving intention diagram in a non-intersection scene determined by a real-time collision detection method based on driving intention and trajectory prediction provided by the application.

[0039] Figure 3 A trajectory prediction output diagram in a non-intersection scene determined by a real-time collision detection method based on driving intention and trajectory prediction provided by the application.

[0040] Figure 4 A driving intention diagram in an intersection scene determined by a real-time collision detection method based on driving intention and trajectory prediction provided by the application.

[0041] Figure 5 A trajectory prediction output diagram in an intersection scene determined by a real-time collision detection method based on driving intention and trajectory prediction provided by the application.

[0042] Figure 6 A collision scene diagram in an intersection after introducing space and time redundancy by a real-time collision detection method based on driving intention and trajectory prediction provided by the application.

[0043] Figure 7 A collision warning output diagram obtained when redundancy is not added in a real-time collision detection method based on driving intention and trajectory prediction provided by the application.

[0044] Figure 8 A collision warning output diagram after adding redundancy in a real-time collision detection method based on driving intention and trajectory prediction provided by the application. DETAILED DESCRIPTION

[0045] The application will be described in detail below with reference to the accompanying drawings and examples.

[0046] The present invention provides a real-time collision detection method based on driving intention and trajectory prediction. The basic idea is to construct a driving intention prediction model based on features extracted from non-intersection scenarios and intersection scenarios, and use a Bayesian optimization algorithm based on tree-shaped Parzen estimation to complete the training of the driving intention prediction model. The trained driving intention prediction model is used to predict the driving intention of obstacle vehicles around a controlled vehicle. The driving trajectories of obstacle vehicles around the controlled vehicle are then predicted for non-intersection scenarios and intersection scenarios respectively. By comparing the driving trajectories of the controlled vehicle and the obstacle vehicle, it is determined whether the controlled vehicle and the obstacle vehicle will collide.

[0047] The present invention provides a real-time collision detection method based on driving intention and trajectory prediction, which specifically includes the following steps:

[0048] Step 1: Obtain vehicle trajectory information and the true value of the vehicle driving intention annotated therein to construct a training sample dataset.

[0049] Specifically, the present invention uses the publicly available datasets NGSIM and INTERACTION as training sample datasets. The NGSIM dataset can be used to train a driving intention prediction model for non-intersection scenarios, while the INTERACTION dataset can be used to train a driving intention prediction model for intersection scenarios. The present invention limits the number of samples in each class to the minimum among all classes to mitigate data imbalance, and divides the training, test, and validation sets into a 7:1.5:1.5 ratio.

[0050] Step 2: Build a driving intention prediction model based on a single-layer LSTM network with a non-deep network. The model structure is as follows: Figure 1 As shown, the driving intentions of obstacle vehicles located around the controlled vehicle are predicted, including turning left, turning right, or going straight.

[0051] The input of the driving intention prediction model constructed by the present invention is shown in the following formula:

[0052]

[0053] Among them, I i is the historical trajectory sequence feature of the i-th obstacle vehicle, is the feature of the i-th obstacle vehicle at time t. The feature of the non-intersection scene is expressed as As shown in the following formula:

[0054]

[0055] in, is the distance from the starting point of the road to the current vehicle position along the road reference line at time tk, is the offset perpendicular to the current position along the road reference line at time tk, is the offset of the obstacle vehicle at time t-k relative to the current heading angle of the controlled vehicle, is the speed value of the obstacle vehicle at time t-k, is the acceleration value of the obstacle vehicle at time t-k.

[0056] The feature representation of the intersection scenario is as shown in the following formula:

[0057]

[0058] wherein, and is the position coordinate of the i-th obstacle vehicle in the plane rectangular coordinate system at time t-k under the given UTM coordinate system, is the speed value of the obstacle vehicle along the X-axis at time t-k, is the speed value of the obstacle vehicle along the Y-axis at time t-k.

[0059] The output of the driving intention prediction model is the occurrence probability of each type of intention corresponding to different scenarios, and the intention with the highest probability among all feasible intentions is taken as the final intention. The predicted driving intention in a non-intersection scenario is as shown in Figure 2 , where positive integer 1 represents that the vehicle changes lane to the left but does not enter the target lane, and positive integer 2 represents that the vehicle changes lane to the left and has entered the target lane; negative integer -1 represents that the vehicle changes lane to the left but does not enter the target lane, and negative integer -2 represents that the vehicle changes lane to the left and has entered the target lane. The predicted driving intention in an intersection scenario is as shown in Figure 4 , wherein positive integer 1 represents that the vehicle has a lane change intention to the left at the intersection, and positive integer 2 represents that the vehicle has a left turn intention at the intersection; negative integer -1 represents that the vehicle has a lane change intention to the left at the intersection, and negative integer -2 represents that the vehicle has a left turn intention at the intersection.

[0060] Step 3: Obtain the best hyperparameters of the driving intention prediction model by using the Bayesian optimization algorithm based on the tree-structured Parzen estimator (TPE), and complete the training of the driving intention prediction model.

[0061] For example, the best hyperparameters of the driving intention prediction model obtained by using the Bayesian optimization algorithm based on TPE are shown in Table 1:

[0062] Table 1 Best hyperparameters of the driving intention prediction model

[0063] TPE-LSTM hyperparameters Non-intersection scenarios Intersection scene full connection layer hidden layer size hs1 30 45 [ LSTM layer hidden size hs2 ] 45 45 Number of LSTM layers ln 1 3 Learning rate α 0.001 0.0001 Number of training rounds ep 260 180 Batch size bs 4096 2048

[0064] Step 4, in actual use, the vehicle trajectory information of the obstacle vehicle is taken as the input of the driving intention prediction model trained, the driving intention of the obstacle vehicle is calculated by the driving intention prediction model, and the driving trajectory of the obstacle vehicle is predicted for the non-intersection scene and the intersection scene respectively, and the trajectory prediction range is a set period, which can be 8 seconds. The vehicle trajectory information includes vehicle position, speed, acceleration and heading angle.

[0065] Specifically, in the non-intersection scene, the trajectory is generated in the S direction (along the road direction) and the L direction (perpendicular to the road direction) of the Frenet coordinate system, including linear driving vehicles and vehicles with lane changing intention, as follows:

[0066] For linear driving vehicles, it is assumed that the vehicle maintains uniform acceleration motion in the S direction and the speed does not exceed the maximum speed threshold, and continuously approaches the center line of the lane in the L direction. The process is represented by the following formula:

[0067]

[0068] where s t is the distance traveled by the obstacle vehicle along the road reference line at time t, s t+1 is the distance traveled by the obstacle vehicle along the road reference line at time t+1, v s is the projection of the speed of the obstacle vehicle in the S direction, a s is the projection of the acceleration of the obstacle vehicle in the S direction, Δt is the prediction time interval, is the L coordinate of the projection point of the obstacle vehicle on the center line at time t, k is the speed of the obstacle vehicle approaching the center line, l t is the offset of the obstacle vehicle along the road reference line perpendicular to the current position at time t, l t+1 is the offset of the obstacle vehicle along the road reference line perpendicular to the current position at time t+1.

[0069] For vehicles with lane changing intention, fourth order polynomial s(t) and fifth order polynomial l(t) are used to model the trajectories in the S direction and the L direction, respectively, as shown in the following formula:

[0070]

[0071] This set of equations describes the solution process for the two state variables s(t) and l(t) and their first and second derivatives. s(t) is the distance traveled by the obstacle vehicle along the road reference line, and l(t) is the distance traveled by the obstacle vehicle along the direction perpendicular to the road reference line. s(0) = s0 represents the state of s0 substituted into the first expression for s(t), represents the substitution of The first expression of the state construction is substituted into s(0) to solve the second expression of s(t), and the coefficients of the polynomial of s(t) can be solved by simultaneous equations. The first expression of the state construction is substituted into s(0) to solve the second expression of s(t), and the coefficients of the polynomial of s(t) can be solved by simultaneous equations. The first expression of the state construction is substituted into s(0) to solve the second expression of s(t), and the coefficients of the polynomial of s(t) can be solved by simultaneous equations. s0, l0 and their higher-order derivatives can be calculated according to the current state of the obstacle vehicle, T1 is the lane changing time, and the candidate lane changing trajectory of the obstacle vehicle can be calculated in the Frenet coordinate system by sampling T1. For the above plurality of candidate lane changing trajectories, the optimal trajectory is selected by using the cost function shown in the following formula:

[0072] cost lc (T1)=w1T1+w2max{a n} (6)

[0073] wherein w1 and w2 are weight parameters, and max{a n} is the maximum normal acceleration in the trajectory, and the trajectory with the minimum cost function value is selected as the lane changing trajectory.

[0074] In the intersection scenario, the third polynomial x(t) and y(t) are used to model the turning or straight trajectory in the X and Y directions of the UTM coordinate system as shown in the following formula:

[0075]

[0076] The first expression of the state construction is substituted into s(0) to solve the second expression of s(t), and the coefficients of the polynomial of s(t) can be solved by simultaneous equations. The first expression of the state construction is substituted into s(0) to solve the second expression of s(t), and the coefficients of the polynomial of s(t) can be solved by simultaneous equations. The first expression of the state construction is substituted into s(0) to solve the second expression of s(t), and the coefficients of the polynomial of s(t) can be solved by simultaneous equations. The first expression of the state construction is substituted into s(0) to solve the second expression of s(t), and the coefficients of the polynomial of s(t) can be solved by simultaneous equations. The first expression of the state construction is substituted into s(0) to solve the second expression of s(t), and the coefficients of the polynomial of s(t) can be solved by simultaneous equations. T1 is the time required for the obstacle vehicle to pass through the intersection, and x exit and y exit are the UTM coordinates of the first point on the center line of the exit lane. A series of candidate trajectories are obtained by sampling T1, and the optimal trajectory in the intersection scenario is calculated by using formula (6). In the present application, the longest predicted time of the trajectory is 8 seconds, and when the predicted time is less than 8 seconds, the method of extending the straight trajectory by using formula (4) is used to extend the trajectory to 8 seconds.

[0077] The driving trajectory predicted by the present application in the non-intersection scenario is shown in Figure 3 , and the driving trajectory in the intersection scenario is shown in Figure 5 .

[0078] Step 5, determine the oriented bounding box (OBB) of the controlled vehicle and the obstacle vehicle related thereto, and expand the length and width of the determined OBB by k times to set the collision space redundancy; after determining the OBB of the controlled vehicle and the obstacle vehicle, introduce the time difference of collision (TDC) to represent the time interval of the controlled vehicle and the obstacle vehicle reaching the same position, if the time difference of collision of the two is less than the set threshold δ, it is determined that the two vehicles will collide, otherwise it is determined that the two vehicles will not collide.

[0079] The existing collision detection algorithm usually only evaluates whether the OBB of the two trajectories intersects at the same time. However, due to the inevitable errors in path and speed prediction, the reliability of the algorithm is affected. In order to solve this problem, the present application enhances the reliability of the traditional collision detection algorithm by introducing spatial and temporal redundancy. Specifically, by expanding the length and width of the OBB by k times, spatial redundancy is achieved, which can improve the detection accuracy to a certain extent; by introducing the time difference of collision TDC, temporal redundancy is achieved, that is, if the value of the time difference of collision is less than the threshold, it is considered that the two vehicles will collide on the future space-time trajectory.

[0080] Intuitively, the accuracy of the predicted trajectory will deviate in position and speed over time. Therefore, when designing spatial and temporal redundancy, the present application sets the expansion factor k of the OBB and the TDC threshold δ t , and the values of the two variables will increase over time, so the present application designs a linear saturation function r(t) to represent the redundancy, as shown in the following formula:

[0081] r(t)=max{at,r max} (8)

[0082] where a is the linear coefficient, r max is the maximum value of the redundancy, and max{at,r max} takes the larger value of at and r max . As Figure 6 shown, after introducing the redundancy judgment, taking the intersection scene as an example, the ego vehicle can stably output the collision point and achieve early perception, greatly optimizing the problem of insufficient collision danger perception of other collision algorithms. The collision warning output without joining the redundancy is shown in Figure 7 , where 0 represents no collision in the current frame, and 1 represents collision in the current frame. After joining the redundancy setting, as Figure 8As shown, 0 represents no collision in the current frame, and 1 represents collision in the current frame, the perception of the controlled vehicle to the collision of the surrounding obstacle vehicle is more stable, and the problem of frequent collision warning mutation at the collision critical point can be avoided.

[0083] To sum up, the above is only the preferred embodiment of the present application, not for limiting the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A real-time collision detection method based on driving intention and trajectory prediction, characterized in that: The specific steps include: Step 1: Obtain vehicle trajectory information and the true value of the vehicle driving intention annotated therein to construct a training sample dataset; Step 2: A driving intention prediction model is established based on a single-layer LSTM network of a non-deep network to predict the driving intention of obstacle vehicles around the controlled vehicle. The input of the driving intention prediction model is shown in the following formula: Among them, I i is the historical trajectory sequence feature of the i-th obstacle vehicle, is the feature of the i-th obstacle vehicle at time t; the feature of the non-intersection scene is expressed as As shown in the following formula: in, is the distance from the starting point of the road to the current vehicle position along the road reference line at time tk, is the offset perpendicular to the current position along the road reference line at time tk, is the offset of the obstacle vehicle relative to the current heading angle of the controlled vehicle at time tk, is the speed of the obstacle vehicle at time tk, is the acceleration value of the obstacle vehicle at time tk; The feature representation of the intersection scene is As shown in the following formula: in, and is the position coordinate of the i-th obstacle vehicle in the plane rectangular coordinate system at time tk in the given UTM coordinate system, is the speed of the obstacle vehicle along the X axis at time tk, is the speed of the obstacle vehicle along the Y axis at time tk; Step 3: Use the Bayesian optimization algorithm based on tree-shaped Parzen estimation to obtain the optimal hyperparameters of the driving intention prediction model and complete the training of the driving intention prediction model; Step 4: In actual use, the vehicle trajectory information of the obstructing vehicle is used as the input of the trained driving intention prediction model. The driving intention prediction model calculates the driving intention of the obstructing vehicle and then predicts the driving trajectory of the obstructing vehicle for non-intersection scenarios and intersection scenarios respectively. The trajectory prediction range is the set time period; Step 5: Determine the directional bounding boxes of the controlled vehicle and its associated obstacle vehicles, and expand the length and width of the directional bounding boxes by a factor of k. Then, introduce the collision time difference to represent the time interval between the controlled vehicle and the obstacle vehicle reaching the same position. If the collision time difference between the two is less than the set threshold δ, it is determined that the two vehicles will collide; otherwise, it is determined that the two vehicles will not collide.

2. The real-time collision detection method according to claim 1, characterized in that: The prediction method for the driving trajectory of vehicles with obstacles in non-intersection scenarios is as follows: For a vehicle traveling in a straight line, assuming that it maintains uniform acceleration along the road direction and its speed does not exceed the maximum speed threshold, and continuously approaches the lane centerline in the direction perpendicular to the road, the process can be expressed as shown in the following formula: Among them, s t is the distance traveled by the obstacle vehicle along the road reference line at time t, s t+1 is the distance traveled by the obstacle vehicle along the road reference line at time t+1, v s is the projection of the obstacle vehicle’s velocity in the S direction, a s is the projection of the obstacle vehicle’s acceleration in the S direction, Δt is the prediction time interval, is the L coordinate of the projection point of the obstacle vehicle on the center line at time t, k is the speed of the obstacle vehicle approaching the center line, l t is the offset of the obstacle vehicle along the road reference line perpendicular to the current position at time t, l t+1 The offset of the obstacle vehicle along the road reference line perpendicular to the current position at time t+1; For vehicles with lane-changing intentions, the fourth-order polynomial s(t) and the fifth-order polynomial l(t) are used to model the trajectories in the S and L directions, as shown in the following equations: s(t) is the distance traveled by the obstructing vehicle along the road reference line, l(t) is the distance traveled by the obstructing vehicle in the direction perpendicular to the road reference line; s(0) = s0 means that the state of s0 is substituted to construct the first expression for solving s(t). Indicates substitution The state of constructing the second expression of solving s(t), and so on to solve the coefficients of the s(t) polynomial; l(0) = l0 means substituting the state of l0 to construct the first expression, Indicates substitution The second expression is constructed based on the state of , and the coefficients of the l(t) polynomial are solved in this way. s0, l0 and their higher-order derivatives are all calculated from the current state of the obstacle vehicle. T1 is the lane change time. The candidate lane change trajectory of the obstacle vehicle is calculated by sampling T1 in the Frenet coordinate system. The optimal trajectory is then selected using the cost function shown in the following formula: cost lc (T1)=w1T1+w2max{a n } Among them, w1 and w2 are weight parameters, max{a n } is the maximum normal acceleration in the trajectory, and the trajectory with the minimum cost function value is selected as the lane change trajectory.

3. The real-time collision detection method according to claim 1, characterized in that: The method for predicting the driving trajectory of an obstacle vehicle in an intersection scenario is as follows: The turning or straight trajectory in the X and Y directions in the UTM coordinate system is modeled as follows: Among them, x(0)=x0 means substituting the state of x0 to construct the first expression of x(t). Indicates substitution The state of l0 is used to construct the second expression for solving x(t), and so on, the coefficients of the polynomial x(t) can be solved; y(0) = y0 means substituting the state of l0 to construct the first expression, Indicates substitution The state of the second expression is constructed, and so on, the coefficients of the y(t) polynomial are solved simultaneously; T1 is the time required for the obstacle vehicle to pass through the intersection, x exit and y exit is the UTM coordinate of the first point on the center line of the exit lane; a series of candidate trajectories are obtained by sampling T1, and the trajectory with the smallest cost function value is selected as the optimal trajectory.

4. The real-time collision detection method according to claim 1, characterized in that: The NGSIM dataset is used to train the driving intention prediction model for non-intersection scenarios, and the INTERACTION dataset is used to train the driving intention prediction model for intersection scenarios.

5. The real-time collision detection method according to claim 1, characterized in that: The driving intention includes turning left, turning right, or going straight.

6. The real-time collision detection method according to claim 1, characterized in that: The set period is 8 seconds.

7. The real-time collision detection method according to claim 1, characterized in that: The vehicle trajectory information includes vehicle position, speed, acceleration and heading angle.

Images