Vehicle motion trail prediction method and system

Abstract

The embodiment of the invention discloses a vehicle motion track prediction method and system, and the method comprises the steps: constructing a topological structure chart according to map information, carrying out the traversing of the topological structure chart, obtaining a drivable lane sequence around a target vehicle, and carrying out the fitting of a trend track for each target vehicle through employing the perception information of the target vehicle, and calculating the cost from the target vehicle to each drivable lane sequence of the target vehicle by combining the historical trajectory, and for each target vehicle, finally generating a preset number of prediction trajectories with the maximum probability of the target vehicle by using the cost. The prediction method is a prediction scheme based on rules and structures, the algorithm consumption is low, the deployment is simple, and the prediction requirement of the actual trajectory can be met. In addition, the topological structure chart is constructed in combination with priori knowledge such as map information and topological relation, and the kinematics prediction accuracy can be improved; in addition, the prediction trajectory is generated based on the probability, and the distribution characteristics of the prediction trajectory can be qualitatively and quantitatively reflected.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and more specifically, to a vehicle motion trajectory prediction method and system. Background Art

[0002] Autonomous driving technology is one of the most cutting-edge research directions in today's automotive industry and artificial intelligence. With the development of technology, autonomous vehicles (AVs) have shown great potential in improving road safety, alleviating traffic congestion, and improving travel efficiency. However, in order to achieve these goals, the autonomous driving system needs to be highly intelligent and able to accurately predict and respond to changes in the surrounding environment, including the behavior of other vehicles, pedestrians, bicycles and other traffic participants.

[0003] There are two main solutions for vehicle motion trajectory prediction in the existing technology. The first one is to predict the trajectory based on a deep learning training model. This prediction solution needs to collect a large amount of data to train the model, and the amount of calculation required for this solution is large. The other is a rule-based prediction solution, which is mainly based on vehicle kinematics, that is, prediction based on the vehicle's motion state, and its accuracy needs to be improved. Summary of the invention

[0004] The present application provides a vehicle motion trajectory prediction method and system to improve the efficiency and accuracy of predicting vehicle trajectory. The specific scheme is as follows:

[0005] In a first aspect, the present application provides a vehicle motion trajectory prediction method, comprising:

[0006] Acquire map information upstream of the self-driving vehicle and perception information of at least one target vehicle, and convert the map information into a topological structure diagram;

[0007] Among them, the map information includes static geometric elements of the lane, and there is a topological connection relationship between the static geometric elements, and the static geometric elements include lane lines and edge lines; the perception information includes size, type, location and driving information, and the driving information includes speed, acceleration and heading angle.

[0008] Using a depth-first search algorithm to search in the topological structure graph, a sequence of drivable lanes around each target vehicle is obtained;

[0009] For each target vehicle, the historical trajectory is fitted using the perception information of the target vehicle to obtain the trend trajectory of the target vehicle;

[0010] For each target vehicle, the cost of the target vehicle to each drivable lane sequence is calculated using the historical trajectory and trend trajectory of the target vehicle;

[0011] For each target vehicle, a preset number of predicted trajectories with the highest probability for the target vehicle are generated using the cost of the target vehicle to each corresponding drivable lane sequence.

[0012] Optionally, converting the map information into a topological structure diagram includes:

[0013] The connection points between the static geometric elements are taken as points, and the topological connection relationships between the static geometric elements are constructed as connection edges to obtain a topological structure graph.

[0014] Optionally, for each target vehicle, fitting its historical trajectory using the perception information of the target vehicle to obtain the trend trajectory of the target vehicle includes:

[0015] For each target vehicle, the historical trajectory is fitted using the perception information of the target vehicle;

[0016] The historical trajectory is fitted using a least squares polynomial to obtain the trend trajectory of the target vehicle.

[0017] Optionally, for each target vehicle, using the historical trajectory and trend trajectory of the target vehicle to calculate the cost of the target vehicle to each corresponding drivable lane sequence includes:

[0018] For each target vehicle, the historical trajectory and trend trajectory of the target vehicle are used to calculate the cost of the target vehicle to each corresponding drivable lane sequence using a cost calculation formula; the cost calculation formula is expressed as:

[0019]

[0020] In the formula, cost is the cost of the target vehicle to each drivable lane sequence, d history is the distance from the target vehicle to the corresponding drivable lane sequence in the historical trajectory, d trend is the distance from the target vehicle to the corresponding drivable lane sequence in the trend trajectory, w1, w2, w3 and w4 are all weight parameters, v lat is the lateral speed and heading is the heading angle.

[0021] Optionally, for each target vehicle, using the cost of the target vehicle to each corresponding drivable lane sequence to generate a preset number of predicted trajectories with the highest probability for the target vehicle includes:

[0022] For each target vehicle, the probability of the target vehicle going to each drivable lane sequence is calculated using the cost of the target vehicle going to each drivable lane sequence.

[0023] According to the probability of the target vehicle corresponding to each drivable lane sequence and the vehicle kinematic model, a preset number of predicted trajectories with the highest probability for the target vehicle are generated.

[0024] Optionally, the calculating the probability of the target vehicle reaching each drivable lane sequence using the cost of the target vehicle reaching each drivable lane sequence includes:

[0025] The cost of the target vehicle to each drivable lane sequence is used, and the probability calculation formula is used to calculate the probability of the target vehicle to each drivable lane sequence; the probability calculation formula is expressed as:

[0026]

[0027] In the formula, p i represents the probability of the target vehicle reaching the i-th drivable lane sequence, N represents the total number of drivable lane sequences, i represents the sequence number of the drivable lane sequence, and cost i Represents the cost of the target vehicle to the i-th drivable lane sequence.

[0028] Optionally, the generating a preset number of predicted trajectories with the highest probability of the target vehicle according to the probability of the target vehicle to each drivable lane sequence in combination with the vehicle kinematic model comprises:

[0029] Determine the maximum probability and the second maximum probability among the probabilities of the target vehicle in all drivable lane sequences;

[0030] Select the two drivable lane sequences corresponding to the maximum probability and the second maximum probability;

[0031] Set the predicted trajectory length and the interval between predicted trajectory points;

[0032] The fitting curve is used in combination with the vehicle kinematics model to fit the predicted trajectory of the target vehicle to the two drivable lane sequences according to the predicted trajectory length and the interval between the predicted trajectory points.

[0033] Optionally, the fitting curve adopts a Bezier curve, and the predicted trajectory is expressed as

[0034] B(n)=P0(1-n) 4 +4P1n(1-n) 3 +6P2n 2 (1-n) 2 +4P3n 3 (1-n)

[0035] +P4n 4 , n∈[0,1]

[0036] Wherein, B(n) represents the predicted trajectory, P0 represents the control point located at the center of the front of the target vehicle, P1 represents the position point where the target vehicle moves along the trend trajectory for the first preset time, P2 is the position point where the target vehicle moves along the center line of the lane for the second preset time after being projected onto the center line of the lane, and P3 and P4 are the position points where the target vehicle continues to move along the center line of the lane for the first preset time and the second preset time, respectively.

[0037] In a second aspect, the present application provides a vehicle motion trajectory prediction system, including a map information processing module, an intention generation module and a trajectory generation module;

[0038] The map information processing module is configured to obtain map information upstream of the self-driving vehicle and perception information of at least one target vehicle, and convert the map information into a topological structure graph; use a depth-first search algorithm to search in the topological structure graph to obtain a sequence of drivable lanes around each target vehicle;

[0039] The intention generation module is configured to, for each target vehicle, use the perception information of the target vehicle to fit its historical trajectory to obtain the trend trajectory of the target vehicle; for each target vehicle, use the historical trajectory and trend trajectory of the target vehicle to calculate the cost of the target vehicle to each corresponding drivable lane sequence;

[0040] The trajectory generation module is configured to generate, for each target vehicle, a preset number of predicted trajectories with the highest probability for the target vehicle using the cost of the target vehicle to each corresponding drivable lane sequence.

[0041] Optionally, the map information processing module is specifically configured to: use the connection points between the static geometric elements as points, and construct the topological connection relationships between the static geometric elements as connection edges to obtain a topological structure diagram.

[0042] Optionally, the intention generation module is specifically configured to: for each target vehicle, fit the historical trajectory using the perception information of the target vehicle; and fit the historical trajectory using a least squares polynomial to obtain the trend trajectory of the target vehicle.

[0043] Optionally, the intention generation module is specifically configured as follows:

[0044] For each target vehicle, the historical trajectory and trend trajectory of the target vehicle are used to calculate the cost of the target vehicle to each corresponding drivable lane sequence using a cost calculation formula; the cost calculation formula is expressed as:

[0045]

[0046] In the formula, cost is the cost of the target vehicle to each drivable lane sequence, d history is the distance from the target vehicle to the corresponding drivable lane sequence in the historical trajectory, d trend is the distance from the target vehicle to the corresponding drivable lane sequence in the trend trajectory, w1, w2, w3 and w4 are all weight parameters, v lat is the lateral speed and heading is the heading angle.

[0047] Optionally, the trajectory generation module is specifically configured as follows:

[0048] For each target vehicle, the cost of the target vehicle to each corresponding drivable lane sequence is used to calculate the probability of the target vehicle to each corresponding drivable lane sequence; based on the probability of the target vehicle to each corresponding drivable lane sequence combined with the vehicle kinematic model, a preset number of predicted trajectories with the highest probability for the target vehicle are generated.

[0049] Optionally, the trajectory generation module is specifically configured as follows:

[0050] The cost of the target vehicle to each drivable lane sequence is used, and the probability calculation formula is used to calculate the probability of the target vehicle to each drivable lane sequence; the probability calculation formula is expressed as:

[0051]

[0052] In the formula, p i represents the probability of the target vehicle reaching the i-th drivable lane sequence, N represents the total number of drivable lane sequences, i represents the sequence number of the drivable lane sequence, and cost i Represents the cost of the target vehicle to the i-th drivable lane sequence.

[0053] Optionally, the trajectory generation module is specifically configured as follows:

[0054] Determine the maximum probability and the second maximum probability among the probabilities of the target vehicle in all drivable lane sequences;

[0055] Select the two drivable lane sequences corresponding to the maximum probability and the second maximum probability;

[0056] Set the predicted trajectory length and the interval between predicted trajectory points;

[0057] The fitting curve is combined with the vehicle kinematic model to fit the predicted trajectory of the target vehicle to the two drivable lane sequences according to the predicted trajectory length and the interval between the predicted trajectory points. The fitting curve adopts a Bezier curve, and the predicted trajectory is expressed as

[0058] B(n)=P0(1-n) 4+4P1n(1-n) 3 +6P2n 2 (1-n) 2 +4P3n 3 (1-n)

[0059] +P4n 4 , n∈[0,1]

[0060] Wherein, B(n) represents the predicted trajectory, P0 represents the control point located at the center of the front of the target vehicle, P1 represents the position point where the target vehicle moves along the trend trajectory for the first preset time, P2 is the position point where the target vehicle moves along the center line of the lane for the second preset time after being projected onto the center line of the lane, and P3 and P4 are the position points where the target vehicle continues to move along the center line of the lane for the first preset time and the second preset time, respectively.

[0061] The innovative features of the embodiments of the present application include:

[0062] The embodiment of the present application discloses a vehicle motion trajectory prediction method and system, which constructs a topological structure diagram based on map information, thereby traversing the topological structure diagram to obtain a sequence of drivable lanes around a target vehicle, and then for each target vehicle, the target vehicle's perception information is used to fit a trend trajectory, and then the cost of the target vehicle to each of its own drivable lane sequences is calculated in combination with the historical trajectory. For each target vehicle, the cost is used to finally generate a preset number of predicted trajectories with the highest probability for the target vehicle. The prediction method of the present application is a prediction scheme based on rules and structures, with low algorithm consumption and simple deployment, which can meet the prediction needs of actual trajectories. In addition, the present application combines map information and prior knowledge such as topological relationships to construct a topological structure diagram, which can improve the accuracy of kinematic predictions; in addition, the present application generates predicted trajectories based on probability, which can qualitatively and quantitatively reflect the distribution characteristics of the predicted trajectories. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art description are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative work.

[0064] Figure 1 A schematic diagram of a process flow of a vehicle motion trajectory prediction method provided in an embodiment of the present application;

[0065] Figure 2 A typical lane topology diagram provided for an embodiment of the present application;

[0066] Figure 3A schematic diagram of a process for fitting a trend trajectory of a target vehicle provided in an embodiment of the present application;

[0067] Figure 4 A schematic diagram of a trend trajectory of a target vehicle provided in an embodiment of the present application;

[0068] Figure 5 A schematic diagram of a process for generating a preset number of predicted trajectories provided in an embodiment of the present application;

[0069] Figure 6 A schematic diagram of using a Bezier curve to select control points provided in an embodiment of the present application;

[0070] Figure 7 A schematic diagram of the structure of a vehicle motion trajectory prediction system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0071] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0072] It should be noted that the terms "including" and "having" and any variations thereof in the embodiments of the present application and the accompanying drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or devices.

[0073] Based on the defects of the prior art, it is particularly important to propose a method that can accurately predict the motion trajectory of an autonomous vehicle. This application proposes a vehicle motion trajectory prediction method and system that can analyze data from vehicle sensors, including radar, laser radar (LiDAR), cameras, etc., as well as possible perception information such as road conditions, traffic signals, etc. Through real-time processing and analysis of these data, the system can predict the possible motion trajectory of the vehicle in the future.

[0074] The embodiment of the present application discloses a vehicle motion trajectory prediction method and system, which constructs a topological structure diagram based on map information, thereby traversing the topological structure diagram to obtain a sequence of drivable lanes around a target vehicle, and then for each target vehicle, the target vehicle's perception information is used to fit a trend trajectory, and then the cost of the target vehicle to each of its own drivable lane sequences is calculated in combination with the historical trajectory. For each target vehicle, the cost is used to finally generate a preset number of predicted trajectories with the highest probability for the target vehicle. The prediction method of the present application is a prediction scheme based on rules and structures, with low algorithm consumption and simple deployment, which can meet the prediction needs of actual trajectories. In addition, the present application combines map information and prior knowledge such as topological relationships to construct a topological structure diagram, which can improve the accuracy of kinematic predictions; in addition, the present application generates predicted trajectories based on probability, which can qualitatively and quantitatively reflect the distribution characteristics of the predicted trajectories.

[0075] The embodiments of the present application are described in detail below.

[0076] Figure 1 A schematic diagram of a vehicle motion trajectory prediction method provided in an embodiment of the present application is shown. Figure 1 As shown, the purpose of this application is to predict the running trajectory of each target vehicle.

[0077] The present application provides a vehicle motion trajectory prediction method, comprising:

[0078] S1, obtaining map information upstream in the driving direction of the self-driving vehicle and perception information of at least one target vehicle, and converting the map information into a topological structure diagram;

[0079] Among them, the map information includes static geometric elements of the lane, and there is a topological connection relationship between the static geometric elements, and the static geometric elements include lane lines and edge lines; the perception information includes size, type, location and driving information, and the driving information includes speed, acceleration and heading angle.

[0080] It is worth noting that a conventional topological structure diagram should include edges and points, where the edges represent the geometric shape of the diagram and the points represent the connection relationship of the edges. Therefore, the present application learns that the topological connection relationship of static geometric elements can construct a topological structure diagram. The specific process includes: taking the connection points between the static geometric elements as points, and constructing the topological connection relationship between the static geometric elements as connecting edges to obtain a topological structure diagram.

[0081] S2, searching the topological structure graph using a depth-first search algorithm to obtain a sequence of drivable lanes around the target vehicle;

[0082] refer to Figure 2 , Figure 2is a typical lane topology diagram provided in the embodiment of the present application. Figure 2 In the example, each lane is numbered from 1 to 11. The present application uses a depth-first-search (DFS) algorithm to traverse a total of four feasible lane sequences, namely {9, 5, 1}, {10, 6, 2}, {11, 7, 3}, and {11, 8, 4}.

[0083] S3, for each target vehicle, fitting its historical trajectory using the perception information of the target vehicle to obtain the trend trajectory of the target vehicle;

[0084] In this step, the historical trajectory of each target vehicle can be fitted to obtain a trend trajectory, and the trend trajectory is fitted by the least square method using a polynomial.

[0085] It is worth noting that the perception information also includes driving information, and the present application can obtain the historical trajectory of the target vehicle based on the driving information. The present application can obtain the historical trajectory from a database storing historical trajectories over a period of time (up to 20 seconds), and the historical trajectory includes the above-mentioned target perception information at each moment, and the historical trajectory can be obtained by fitting based on the perception information.

[0086] S4, for each target vehicle, using the historical trajectory and trend trajectory of the target vehicle, calculating the cost of the target vehicle to each corresponding drivable lane sequence;

[0087] In this step, for each target vehicle, the cost of the target vehicle to its own drivable lane sequence can be calculated using the historical trajectory and trend trajectory of the target vehicle. The smaller the cost of the target vehicle to a certain drivable lane sequence, the greater the intention (probability) of the target vehicle to reach the drivable lane.

[0088] S5, for each target vehicle, using the cost of the target vehicle to each corresponding drivable lane sequence, generate a preset number of predicted trajectories with the highest probability for the target vehicle.

[0089] In a specific embodiment of the present application, Figure 3 As shown, for each target vehicle, fitting its historical trajectory using the perception information of the target vehicle to obtain the trend trajectory of the target vehicle includes:

[0090] S31, for each target vehicle, fitting the historical trajectory using the perception information of the target vehicle;

[0091] S32, fitting the historical trajectory using a least squares polynomial to obtain a trend trajectory of the target vehicle.

[0092] Here we take a first-order polynomial as an example:

[0093] y=ax+b

[0094] Where x, y are the position coordinates of the target vehicle, and a, b are the coefficients of the polynomial.

[0095] The trend trajectory obtained by this application is as follows Figure 4 As shown, in Figure 4 The dashed line segment represents the historical trajectory, and the solid line segment represents the fitted trend trajectory. If you want to improve the fitting accuracy of the trend trajectory, you can use a higher-order polynomial to express it.

[0096] In a specific implementation of the present application, for each target vehicle, using the historical trajectory and trend trajectory of the target vehicle, calculating the cost of the target vehicle to each corresponding drivable lane sequence includes:

[0097] For each target vehicle, the historical trajectory and trend trajectory of the target vehicle are used to calculate the cost of the target vehicle to each corresponding drivable lane sequence using a cost calculation formula; the cost calculation formula is expressed as:

[0098]

[0099] In the formula, cost is the cost of the target vehicle to each drivable lane sequence, d history is the distance from the target vehicle to the corresponding drivable lane sequence in the historical trajectory, d trend is the distance from the target vehicle to the corresponding drivable lane sequence in the trend trajectory, w1, w2, w3 and w4 are all weight parameters, v lat is the lateral speed and heading is the heading angle.

[0100] The cost calculation formula of the present application takes into account the distance from the target vehicle to the drivable lane sequence in the historical trajectory and trend trajectory, and the distance can be calculated using the Euclidean distance formula or the traditional distance calculation formula. The smaller the cost of the target vehicle to a drivable lane sequence, the greater the possibility that the target vehicle will reach the drivable lane sequence. The cost calculation formula of the present application also takes into account the lateral speed and heading angle of the target vehicle, which can further improve the accuracy of the calculated cost, thereby improving the accuracy of predicting the vehicle trajectory. Of course, the cost calculation formula of the present application can also take into account factors such as turn signal status, vehicle speed, acceleration, etc. to further improve the accuracy.

[0101] In a specific embodiment of the present application, Figure 5 As shown, for each target vehicle, using the cost of the target vehicle to each corresponding drivable lane sequence, generating a preset number of predicted trajectories with the highest probability for the target vehicle includes:

[0102] S51, for each target vehicle, using the cost of the target vehicle to each corresponding drivable lane sequence, calculate the probability of the target vehicle to each corresponding drivable lane sequence;

[0103] In this step, the cost of the target vehicle to each drivable lane sequence can be used, and the probability calculation formula can be used to calculate the probability of the target vehicle to each drivable lane sequence; the probability calculation formula is expressed as:

[0104]

[0105] In the formula, p i represents the probability of the target vehicle reaching the i-th drivable lane sequence, N represents the total number of drivable lane sequences, i represents the sequence number of the drivable lane sequence, and cost i Represents the cost of the target vehicle to the i-th drivable lane sequence.

[0106] Here, we take the two predicted trajectories Traj1 and Traj2 with the highest generation probability as an example. The corresponding cost1 and cost2 are the two with the lowest cost. The probability of each predicted trajectory is p1 and p2, which are expressed as:

[0107]

[0108]

[0109] For example, assuming that the cost of the target vehicle to the feasible lane sequence A is 10, and the cost of the target vehicle to the feasible lane sequence B is 100, then the probability of the target vehicle to the feasible lane sequence A is The probability of reaching the feasible lane sequence B is

[0110] S52, generating a preset number of predicted trajectories with the highest probability for the target vehicle based on the probability of the target vehicle arriving at each drivable lane sequence and the vehicle kinematic model.

[0111] This step combines the vehicle kinematics model and uses multi-order curve fitting to generate a predicted trajectory. The multi-order curve can use a Bezier curve.

[0112] In a specific implementation of the present application, the generating of a preset number of predicted trajectories with the highest probability of the target vehicle according to the probability of the target vehicle corresponding to each drivable lane sequence and the vehicle kinematic model comprises:

[0113] S521, determining a maximum probability and a second maximum probability among the probabilities of the target vehicle being in all drivable lane sequences;

[0114] S522, selecting two drivable lane sequences corresponding to the maximum probability and the second maximum probability;

[0115] S523, setting the predicted trajectory length and the interval between predicted trajectory points;

[0116] Among them, the application can set the predicted trajectory length to 6 to 8 seconds, each predicted trajectory point has attributes (x, y, t), namely, position and time, and the interval between predicted trajectory points is 0.1 to 0.2 seconds.

[0117] S524, using the fitting curve in combination with the vehicle kinematics model, fitting the predicted trajectory of the target vehicle to the two drivable lane sequences according to the predicted trajectory length and the interval between the predicted trajectory points.

[0118] If the fitting curve of the present application adopts a Bezier curve, the predicted trajectory is expressed as:

[0119] B(n)=P0(1-n) 4 +4P1n(1-n) 3 +6P2n 2 (1-n) 2 +4P3n 3 (1-n)

[0120] +P4n 4 , n∈[0,1]

[0121] Wherein, B(n) represents the predicted trajectory, P0 represents the control point located at the center of the front of the target vehicle, P1 represents the position point where the target vehicle moves along the trend trajectory for the first preset time, P2 is the position point where the target vehicle moves along the center line of the lane for the second preset time after being projected onto the center line of the lane, and P3 and P4 are the position points where the target vehicle continues to move along the center line of the lane for the first preset time and the second preset time, respectively.

[0122] Bezier Curve is used as the fitting curve. The shape of Bezier Curve is mainly composed of control points. The control points are selected as follows Figure 6 As shown:

[0123] like Figure 6As shown in the figure, control point P0 is located at the center of the front of the vehicle, control point P1 is located at the position where the vehicle moves along the trend trajectory for 2 to 3 seconds, control point P2 is located at the position where the vehicle moves along the center line of the lane for 4 to 5 seconds after being projected onto the center line of the lane (dashed line in the figure), and control points P3 and P4 are located at the positions where the vehicle continues to move along the center line of the lane for 2 to 3 and 4 to 5 seconds respectively. The above control point selection method can take into account the vehicle's movement trend (P0, P1) and the driving lane intention (P2, P3, P4), and the continuous Bezier curve points finally formed are the predicted trajectory. If the two predicted trajectories with the highest probability are taken, two predicted trajectories of the target vehicle's movement must be generated separately.

[0124] In a second aspect, the present application provides a vehicle motion trajectory prediction system, such as Figure 7 As shown, it includes a map information processing module 71, an intention generating module 72 and a trajectory generating module 73;

[0125] The map information processing module 71 is configured to obtain map information upstream of the self-driving vehicle and perception information of at least one target vehicle, and convert the map information into a topological structure graph; use a depth-first search algorithm to search in the topological structure graph to obtain a sequence of drivable lanes around each target vehicle;

[0126] Among them, the map information includes static geometric elements of the lane, and there is a topological connection relationship between the static geometric elements, and the static geometric elements include lane lines and edge lines; the perception information includes size, type, location and driving information, and the driving information includes speed, acceleration and heading angle.

[0127] The intention generation module 72 is configured to, for each target vehicle, use the perception information of the target vehicle to fit its historical trajectory to obtain the trend trajectory of the target vehicle; for each target vehicle, use the historical trajectory and trend trajectory of the target vehicle to calculate the cost of the target vehicle to each corresponding drivable lane sequence;

[0128] The trajectory generation module 73 is configured to generate, for each target vehicle, a preset number of predicted trajectories with the highest probability for the target vehicle using the cost of the target vehicle to each corresponding drivable lane sequence.

[0129] Optionally, the map information processing module 71 is specifically configured to: use the connection points between the static geometric elements as points, and construct the topological connection relationships between the static geometric elements as connection edges to obtain a topological structure diagram.

[0130] Optionally, the intention generation module 72 is specifically configured to: for each target vehicle, fit the historical trajectory using the perception information of the target vehicle; and fit the historical trajectory using a least squares polynomial to obtain the trend trajectory of the target vehicle.

[0131] Optionally, the intention generating module 72 is specifically configured to:

[0132] For each target vehicle, the historical trajectory and trend trajectory of the target vehicle are used to calculate the cost of the target vehicle to each corresponding drivable lane sequence using a cost calculation formula; the cost calculation formula is expressed as:

[0133]

[0134] In the formula, cost is the cost of the target vehicle to each drivable lane sequence, d history is the distance from the target vehicle to the corresponding drivable lane sequence in the historical trajectory, d trend is the distance from the target vehicle to the corresponding drivable lane sequence in the trend trajectory, w1, w2, w3 and w4 are all weight parameters, v lat is the lateral speed and heading is the heading angle.

[0135] Optionally, the trajectory generating module 73 is specifically configured as follows:

[0136] For each target vehicle, the cost of the target vehicle to each corresponding drivable lane sequence is used to calculate the probability of the target vehicle to each corresponding drivable lane sequence; based on the probability of the target vehicle to each corresponding drivable lane sequence combined with the vehicle kinematic model, a preset number of predicted trajectories with the highest probability for the target vehicle are generated.

[0137] Optionally, the trajectory generating module 73 is specifically configured as follows:

[0138] The cost of the target vehicle to each drivable lane sequence is used, and the probability calculation formula is used to calculate the probability of the target vehicle to each drivable lane sequence; the probability calculation formula is expressed as:

[0139]

[0140] In the formula, p i represents the probability of the target vehicle reaching the i-th drivable lane sequence, N represents the total number of drivable lane sequences, i represents the sequence number of the drivable lane sequence, and cost i Represents the cost of the target vehicle to the i-th drivable lane sequence.

[0141] Optionally, the trajectory generating module 73 is specifically configured as follows:

[0142] Determine the maximum probability and the second maximum probability among the probabilities of the target vehicle in all drivable lane sequences;

[0143] Select the two drivable lane sequences corresponding to the maximum probability and the second maximum probability;

[0144] Set the predicted trajectory length and the interval between predicted trajectory points;

[0145] The fitting curve is combined with the vehicle kinematic model to fit the predicted trajectory of the target vehicle to the two drivable lane sequences according to the predicted trajectory length and the interval between the predicted trajectory points. The fitting curve adopts a Bezier curve, and the predicted trajectory is expressed as

[0146] B(n)=P0(1-n) 4 +4P1n(1-n) 3 +6P2n 2 (1-n) 2 +4P3n 3 (1-n)

[0147] +P4n 4 , n∈[0,1]

[0148] Wherein, B(n) represents the predicted trajectory, P0 represents the control point located at the center of the front of the target vehicle, P1 represents the position point where the target vehicle moves along the trend trajectory for the first preset time, P2 is the position point where the target vehicle moves along the center line of the lane for the second preset time after being projected onto the center line of the lane, and P3 and P4 are the position points where the target vehicle continues to move along the center line of the lane for the first preset time and the second preset time, respectively.

[0149] The embodiment of the present application discloses a vehicle motion trajectory prediction method and system, which constructs a topological structure diagram based on map information, thereby traversing the topological structure diagram to obtain a sequence of drivable lanes around a target vehicle, and then for each target vehicle, the target vehicle's perception information is used to fit a trend trajectory, and then the cost of the target vehicle to each of its own drivable lane sequences is calculated in combination with the historical trajectory. For each target vehicle, the cost is used to finally generate a preset number of predicted trajectories with the highest probability for the target vehicle. The prediction method of the present application is a prediction scheme based on rules and structures, with low algorithm consumption and simple deployment, which can meet the prediction needs of actual trajectories. In addition, the present application combines map information and prior knowledge such as topological relationships to construct a topological structure diagram, which can improve the accuracy of kinematic predictions; in addition, the present application generates predicted trajectories based on probability, which can qualitatively and quantitatively reflect the distribution characteristics of the predicted trajectories.

[0150] Those skilled in the art can understand that the accompanying drawings are only schematic diagrams of one embodiment, and the modules or processes in the accompanying drawings are not necessarily required to implement the present application.

[0151] Those skilled in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment according to the description of the embodiment, or can be changed accordingly and located in one or more devices different from the embodiment. The modules in the above embodiment can be combined into one module, or can be further divided into multiple sub-modules.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A vehicle motion trajectory prediction method, characterized in that: include: Acquire map information upstream of the self-driving vehicle and perception information of at least one target vehicle, and convert the map information into a topological structure diagram; Using a depth-first search algorithm to search in the topological structure graph, a sequence of drivable lanes around each target vehicle is obtained; For each target vehicle, the historical trajectory is fitted using the perception information of the target vehicle to obtain the trend trajectory of the target vehicle; For each target vehicle, the cost of the target vehicle to each drivable lane sequence is calculated using the historical trajectory and trend trajectory of the target vehicle; For each target vehicle, a preset number of predicted trajectories with the highest probability for the target vehicle are generated using the cost of the target vehicle to each corresponding drivable lane sequence.

2. The vehicle motion trajectory prediction method according to claim 1, characterized in that: The map information includes static geometric elements of the lane, and there is a topological connection relationship between the static geometric elements, and the static geometric elements include lane lines and edge lines; the perception information includes size, type, location and driving information, and the driving information includes speed, acceleration and heading angle.

3. The vehicle motion trajectory prediction method according to claim 2, characterized in that: The converting the map information into a topological structure diagram comprises: The connection points between the static geometric elements are taken as points, and the topological connection relationships between the static geometric elements are constructed as connection edges to obtain a topological structure graph.

4. The vehicle motion trajectory prediction method according to claim 1, characterized in that: For each target vehicle, fitting its historical trajectory using the perception information of the target vehicle to obtain the trend trajectory of the target vehicle includes: For each target vehicle, the historical trajectory is fitted using the perception information of the target vehicle; The historical trajectory is fitted using a least squares polynomial to obtain the trend trajectory of the target vehicle.

5. The vehicle motion trajectory prediction method according to claim 4, characterized in that: For each target vehicle, using the historical trajectory and trend trajectory of the target vehicle, calculating the cost of the target vehicle to each corresponding drivable lane sequence includes: For each target vehicle, the historical trajectory and trend trajectory of the target vehicle are used to calculate the cost of the target vehicle to each corresponding drivable lane sequence using a cost calculation formula; the cost calculation formula is expressed as: In the formula, cost is the cost of the target vehicle to each drivable lane sequence, d history is the distance from the target vehicle to the corresponding drivable lane sequence in the historical trajectory, d trend is the distance from the target vehicle to the corresponding drivable lane sequence in the trend trajectory, w1, w2, w3 and w4 are all weight parameters, v lat is the lateral speed and heading is the heading angle.

6. The vehicle motion trajectory prediction method according to claim 1, characterized in that: The method of generating, for each target vehicle, a preset number of predicted trajectories with the highest probability for the target vehicle by using the cost of the target vehicle to each corresponding drivable lane sequence comprises: For each target vehicle, the probability of the target vehicle going to each drivable lane sequence is calculated using the cost of the target vehicle going to each drivable lane sequence. According to the probability of the target vehicle corresponding to each drivable lane sequence and the vehicle kinematic model, a preset number of predicted trajectories with the highest probability for the target vehicle are generated.

7. The vehicle motion trajectory prediction method according to claim 6, characterized in that: The method of calculating the probability of the target vehicle reaching each drivable lane sequence using the cost of the target vehicle reaching each drivable lane sequence includes: The cost of the target vehicle to each drivable lane sequence is used, and the probability calculation formula is used to calculate the probability of the target vehicle to each drivable lane sequence; the probability calculation formula is expressed as: In the formula, p i represents the probability of the target vehicle reaching the i-th drivable lane sequence, N represents the total number of drivable lane sequences, i represents the sequence number of the drivable lane sequence, and cost i Represents the cost of the target vehicle to the i-th drivable lane sequence.

8. The vehicle motion trajectory prediction method according to claim 6, characterized in that: The method of generating a preset number of predicted trajectories with the highest probability for the target vehicle based on the probability of the target vehicle corresponding to each drivable lane sequence and the vehicle kinematic model comprises: Determine the maximum probability and the second maximum probability among the probabilities of the target vehicle in all drivable lane sequences; Select the two drivable lane sequences corresponding to the maximum probability and the second maximum probability; Set the predicted trajectory length and the interval between predicted trajectory points; The fitting curve is used in combination with the vehicle kinematics model to fit the predicted trajectory of the target vehicle to the two drivable lane sequences according to the predicted trajectory length and the interval between the predicted trajectory points.

9. The vehicle motion trajectory prediction method according to claim 8, characterized in that: The fitting curve adopts a Bezier curve, and the predicted trajectory is expressed as: B(n)=P0(1-n) 4 +4P1n(1-n) 3 +6P2n 2 (1-n) 2 +4P3n 3 (1-n)+P4n 4 ,n∈[0,1] Wherein, B(n) represents the predicted trajectory, P0 represents the control point located at the center of the front of the target vehicle, P1 represents the position point where the target vehicle moves along the trend trajectory for the first preset time, P2 is the position point where the target vehicle moves along the center line of the lane for the second preset time after being projected onto the center line of the lane, and P3 and P4 are the position points where the target vehicle continues to move along the center line of the lane for the first preset time and the second preset time, respectively.

10. A vehicle motion trajectory prediction system, characterized in that: It includes a map information processing module, an intention generation module and a trajectory generation module; The map information processing module is configured to obtain map information upstream of the self-driving vehicle and perception information of at least one target vehicle, and convert the map information into a topological structure graph; use a depth-first search algorithm to search in the topological structure graph to obtain a sequence of drivable lanes around each target vehicle; The intention generation module is configured to fit the historical trajectory of each target vehicle using the perception information of the target vehicle to obtain the trend trajectory of the target vehicle; For each target vehicle, the cost of the target vehicle to each drivable lane sequence is calculated using the historical trajectory and trend trajectory of the target vehicle; The trajectory generation module is configured to generate, for each target vehicle, a preset number of predicted trajectories with the highest probability for the target vehicle using the cost of the target vehicle to each corresponding drivable lane sequence.

Images