Traffic object behavior prediction method, device and equipment and storage medium

Abstract

The application discloses a traffic object behavior prediction method and device, equipment and a storage medium. A target traffic object is selected, and a plurality of points in a first preset range are selected as sensing points with the target traffic object as the center. For reference traffic objects in a second preset range of the target traffic object, the reference traffic objects are taken as field sources of a directed distance field, and the distance and gradient value of each sensing point in the directed distance field are calculated. The distance and gradient value of the sensing point in the directed distance field are converted into an influence factor of the influence of the reference traffic objects on the behavior of the target traffic object. Based on the influence factors of all the reference traffic objects on the target traffic object, the interaction characteristics of the target traffic object and all the reference traffic objects are determined, and the behavior of the target traffic object is predicted based on the interaction characteristics, so that the prediction accuracy can be improved, and the interaction prediction ability with high generalization is learned.

Description

Technical Field

[0001] The present invention relates to autonomous driving technology, and in particular to a method, device, equipment and storage medium for predicting traffic object behavior. Background Art

[0002] Autonomous driving technology uses sensors to sense its surroundings, collects environmental information, and accurately calculates and analyzes the environmental information through a pre-trained unmanned driving model in the control device (i.e., the on-board intelligent brain). Finally, it sends instructions to the ECU (Electronic Control Unit) to control different devices in the autonomous driving vehicle separately, thereby realizing fully automatic operation of the vehicle and achieving the purpose of driving.

[0003] In autonomous driving technology, in order to better predict the behavior of traffic objects and make correct decisions, it is necessary to model the interactions between traffic objects and extract interaction features.

[0004] Most existing behavior prediction solutions are based on deep neural networks, transforming the problem of modeling the interactions between traffic objects into a problem of extracting and fusing their features. However, the features extracted by deep neural networks contain information that is unobservable to the outside world, such as the intentions of the traffic objects. This makes them ineffective in learning the interaction mechanisms between traffic objects and accurately predicting their behavior. Summary of the Invention

[0005] The present invention provides a traffic object behavior prediction method, device, equipment and storage medium, which can improve prediction accuracy and learn interactive prediction capabilities with high generalization.

[0006] In a first aspect, the present invention provides a method for predicting traffic object behavior, comprising:

[0007] A target traffic object is selected, and a plurality of points within a first preset range with the target traffic object as the center are selected as sensing points;

[0008] For a reference traffic object within a second preset range of the target traffic object, using the reference traffic object as a field source of a directed distance field, calculating a distance and a gradient value of each of the sensing points in the directed distance field;

[0009] Converting the distance and gradient value of the perception point in the directed distance field into an influence factor of the reference traffic object on the behavior of the target traffic object;

[0010] Determining interaction characteristics between the target traffic object and all reference traffic objects based on influence factors of all reference traffic objects on the target traffic object;

[0011] The behavior of the target traffic object is predicted based on the interaction feature.

[0012] Optionally, with the target traffic object as the center, several points within a first preset range are selected as sensing points, including:

[0013] Determining an envelope box of the target traffic object;

[0014] Four vertices on the envelope box are selected as sensing points.

[0015] Optionally, taking the reference traffic object as the field source of the directed distance field, calculating the distance and gradient value of each perception point in the directed distance field includes:

[0016] Taking the reference traffic object as a field source of a directed distance field, constructing a distance function from a point in the directed distance field to the field source;

[0017] Substituting the coordinates of the perception point into the distance function to calculate the distance of the perception point in the directed distance field;

[0018] Calculate the gradient value of the distance function at the perception point.

[0019] Optionally, converting the distance and gradient value of the perception point in the directed distance field into an influence factor of the reference traffic object on the behavior of the target traffic object includes:

[0020] Transforming the distance and the gradient value using a preset monotonically decreasing and differentiable transformation function to transform the distance and the gradient value into a first influencing factor and a second influencing factor;

[0021] When only the reference traffic object moves in the global coordinate system, a time partial derivative of the first influencing factor of the sensing point in the global coordinate system is calculated as a third influencing factor;

[0022] In consideration of the relative motion between the reference traffic object and the target traffic object in the global coordinate system, a time total derivative of the first influencing factor of the perception point in the global coordinate system is calculated as a fourth influencing factor.

[0023] Optionally, using a preset monotonically decreasing and differentiable transformation function to transform the distance and the gradient value to transform the distance and the gradient value into a first influencing factor and a second influencing factor includes:

[0024] Substituting the distance into the transformation function to obtain a first impact factor;

[0025] calculating a derivative of the transformation function at the distance;

[0026] The product of the derivative and the gradient is calculated to obtain the second influencing factor.

[0027] Optionally, determining the interaction characteristics between the target traffic object and all reference traffic objects based on the influence factors of all reference traffic objects on the target traffic object includes:

[0028] For each of the reference traffic objects, combining the influencing factor and an attribute vector representing an intrinsic attribute of the reference traffic object into a feature vector;

[0029] Extracting high-dimensional features from each of the feature vectors;

[0030] Aggregate the high-dimensional features of all the feature vectors to obtain interaction features between the target traffic object and all reference traffic objects.

[0031] Optionally, aggregating high-dimensional features of all the feature vectors to obtain interaction features between the target traffic object and all reference traffic objects includes:

[0032] Combining the high-dimensional features of all the feature vectors into a feature matrix;

[0033] A pooling operation is performed on multiple elements of each dimension in the feature matrix to obtain interaction features between the target traffic object and all reference traffic objects.

[0034] Optionally, the pooling operation includes maximum pooling, minimum pooling and average pooling.

[0035] In a second aspect, the present invention further provides a traffic object behavior prediction device, comprising:

[0036] a sensing point determination module, configured to select a target traffic object and, with the target traffic object as the center, select a plurality of points within a first preset range as sensing points;

[0037] a calculation module, configured to calculate the distance and gradient value of each of the sensing points in the directed distance field with respect to a reference traffic object within a second preset range of the target traffic object, using the reference traffic object as a field source of the directed distance field;

[0038] a conversion module, configured to convert the distance and gradient value of the perception point in the directed distance field into an influence factor of the reference traffic object on the behavior of the target traffic object;

[0039] an interaction feature determination module, configured to determine interaction features between the target traffic object and all reference traffic objects based on influence factors of all reference traffic objects on the target traffic object;

[0040] A behavior prediction module is used to predict the behavior of the target traffic object based on the interaction characteristics.

[0041] In a third aspect, the present invention further provides an electronic device, comprising:

[0042] one or more processors;

[0043] a memory for storing one or more programs;

[0044] When the one or more programs are executed by the one or more processors, the one or more processors implement the traffic object behavior prediction method provided by the first aspect of the present invention.

[0045] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the traffic object behavior prediction method provided in the first aspect of the present invention.

[0046] The traffic object behavior prediction method provided by the present invention includes: selecting a target traffic object, and selecting a plurality of points within a first preset range with the target traffic object as the center as perception points. For reference traffic objects within a second preset range of the target traffic object, using the reference traffic object as the source of a directed distance field, calculating the distance and gradient value of each perception point in the directed distance field, converting the distance and gradient value of the perception point in the directed distance field into an influence factor of the reference traffic object's influence on the target traffic object's behavior, determining interaction characteristics between the target traffic object and all reference traffic objects based on the influence factors of all reference traffic objects on the target traffic object, and predicting the target traffic object's behavior based on the interaction characteristics. The present invention uses the reference traffic object as the source of the directed distance field, calculates the distance and gradient value of the perception point of the target traffic object in the directed distance field, converts the distance and gradient value into an influence factor of the reference traffic object's influence on the target traffic object's behavior, and determines the interaction characteristics between the target traffic object and all reference traffic objects based on the influence factors. This method can effectively extract information that is unobservable to the outside world, improving prediction accuracy. The extracted interaction characteristics enable a deep neural network to efficiently characterize and process the interactions of multiple traffic objects, thereby learning highly generalizable interaction prediction capabilities.

[0047] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0049] Figure 1 A flow chart of a method for predicting traffic object behavior provided by an embodiment of the present invention;

[0050] Figure 2 A schematic diagram of the structure of a traffic object behavior prediction device provided by an embodiment of the present invention;

[0051] Figure 3 A schematic structural diagram of an electronic device provided by an embodiment of the present invention.

[0052] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0053] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0054] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0055] Figure 1This is a flow chart of a method for predicting the behavior of a traffic object provided by an embodiment of the present invention. This embodiment is applicable to the case of predicting the behavior of a traffic object. The method can be executed by a traffic object behavior prediction device provided by an embodiment of the present invention. The device can be implemented by software and / or hardware and is usually configured in an electronic device, such as Figure 1 As shown, the traffic object behavior prediction method specifically includes the following steps:

[0056] S101: Select a target traffic object, and select several points within a first preset range centered on the target traffic object as sensing points.

[0057] In an embodiment of the present invention, any traffic participant in the traffic scene is selected as the target traffic object A. For example, the target traffic object A may be the autonomous driving vehicle itself or other traffic participants, such as other vehicles, non-motor vehicles or pedestrians, etc., which is not limited in this embodiment of the present invention.

[0058] After selecting the target traffic object A, a number of points within the first preset range are selected as sensing points with the target traffic object A as the center, and the sensing point set {A i}.

[0059] The pre-steps of traffic object behavior prediction usually involve the identification and detection of traffic objects. After the traffic object is detected, in order to facilitate modeling, the smallest cube that can enclose the traffic object (i.e., envelope) is usually used to represent the traffic object. For example, in some embodiments of the present invention, after selecting the target traffic object A, the envelope of the target traffic object A is determined, and several points on the envelope are selected as perception points A. i For example, for the convenience of selection, the four vertices on the envelope box are selected as the perception points A. i In other embodiments of the present invention, other points on the envelope box may be selected as sensing points, for example, the midpoint of the edge of the envelope box, etc., which is not limited in this embodiment of the present invention.

[0060] S102 : For a reference traffic object within a second preset range of the target traffic object, use the reference traffic object as a source of a directed distance field and calculate the distance and gradient value of each sensing point in the directed distance field.

[0061] In the embodiment of the present invention, the target traffic object A is taken as the center, and other traffic participating objects within the second preset range of the target traffic object A are regarded as reference traffic objects B. j , get a set of reference traffic objects {B j For each reference traffic object B j , with reference to traffic object B jThe signed distance field (SDF) is the source of the signed distance field (SDF), and the distance and gradient value of each sensing point in the signed distance field are calculated. The signed distance field is the distance from the point in space (sensing point) to the reference traffic object B. j The shortest directed distance of the surface is composed, usually, of the reference traffic object B j Internal point to reference traffic object B j The distance to the surface is negative, in reference to traffic object B j External point to reference traffic object B j Distances to surfaces are positive.

[0062] For example, in some embodiments of the present invention, with reference to traffic object B j For the source of the signed distance field, construct the distance function from the point in the signed distance field to the source. After selecting the sensing point, for each sensing point A i , the sensing point A i Substitute the coordinates of into the distance function to calculate the perception point A i The distance in the signed distance field is denoted as SDF(A i ; B j ), at the same time, calculate the distance function at the perception point A i The gradient value at is denoted as grad(SDF(A i ; B j )).

[0063] S103: Convert the distance and gradient value of the perception point in the directed distance field into an influence factor of the reference traffic object on the behavior of the target traffic object.

[0064] In this embodiment of the present invention, the sensing point A i With reference to traffic object B j The distance in the directed distance field can represent the influence of the reference traffic object on the target traffic object, but this influence is in an anti-correlated relationship, that is, the perception point A i With reference to traffic object B j The greater the distance in the directed distance field of the field source, the smaller the influence of the reference traffic object on the target traffic object. Therefore, the distance and gradient value of the sensing point in the directed distance field can be converted into an influence factor of the reference traffic object's influence on the target traffic object's behavior. The influence factor represents the influence of the reference traffic object on the target traffic object and is positively correlated. For example, multiple influence factors can be obtained through various transformations, which are not limited in this embodiment of the present invention.

[0065] For example, in some embodiments of the present invention, the above step S103 includes the following sub-steps:

[0066] 1. Use a preset monotonically decreasing and differentiable transformation function to transform the distance and gradient values, and transform the distance and gradient values ​​into the first influencing factor and the second influencing factor.

[0067] In this embodiment of the present invention, to convert the distance and gradient values, which are inversely correlated with the influence of the reference traffic object on the target traffic object, into influence factors, which are positively correlated with the influence of the reference traffic object on the target traffic object, a preset monotonically decreasing and differentiable transformation function is used to transform the distance and gradient values ​​into a first influence factor and a second influence factor, respectively. In this embodiment of the present invention, considering that the perception point on the target traffic object cannot be located within the reference traffic object, the transformation function can be set to be monotonically decreasing and differentiable in the interval [0, +∞).

[0068] For example, in a specific embodiment of the present invention, the transformation function may be f(x)=exp(-x), which satisfies the condition of monotonically decreasing and differentiable in the interval [0, +∞), and f(x) approaches 0 when the independent variable is positive infinity, which satisfies the condition that the influence of the reference traffic object at infinity on the target traffic object is almost zero. Specifically, the distance SDF(A i ; B j ) is substituted into the transformation function f(x) to obtain the first impact factor f(SDF(A i ; B j )), denoted as s ij . Calculate the transformation function f(x) in SDF(A i ; B j ) derivative f'(SDF(A i ; B j )), calculate the derivative and gradient grad(SDF(A i ; B j )) product f'(SDF(A i ; B j ))*grad(SDF(A i ; B j )), the second impact factor is obtained, denoted as g ij In other embodiments of the present invention, the transformation function may also be other functions that are monotonically decreasing and differentiable in the interval [0, +∞), which is not limited in the embodiment of the present invention.

[0069] 2. When only considering the movement of the reference traffic object in the global coordinate system, the time partial derivative of the first influencing factor of the perception point in the global coordinate system is calculated as the third influencing factor.

[0070] Furthermore, considering that the reference traffic object is in motion, the time factor will also affect the interaction between the reference traffic object and the target traffic object. Therefore, in an embodiment of the present invention, when only considering the movement of the reference traffic object in the global coordinate system, the time partial derivative of the first influencing factor of the perception point in the global coordinate system is calculated as the third influencing factor.

[0071] Specifically, the reference traffic object is represented by an envelope box ABCD, where A, B, C, and D are the vertices of the envelope box. Let l = SDF(Ai; Bj), and the third influencing factor pt ij The calculation formula is as follows:

[0072]

[0073] Among them, V k is the moving speed of point k on the reference traffic object in the global coordinate system, S is the signed distance field determined by the envelope box ABCD, is the gradient of the SDF value at the sampling point Ai to the field source point k.

[0074] 3. Considering the relative motion of the reference traffic object and the target traffic object in the global coordinate system, the time total derivative of the first influencing factor at the perception point in the global coordinate system is calculated as the fourth influencing factor.

[0075] In actual traffic scenarios, both the reference traffic object and the target traffic object are in motion, and both will affect the behavior of the target traffic object. Therefore, in the embodiment of the present invention, considering the relative motion of the reference traffic object and the target traffic object in the global coordinate system, the time total derivative of the first influencing factor at the perception point in the global coordinate system is calculated as the fourth influencing factor. Specifically, the fourth influencing factor dt ij The calculation formula is as follows:

[0076]

[0077] Where V is the moving speed of the target traffic object in the global coordinate system, is the gradient value grad(SDF(A i ; B j )).

[0078] S104: Determine interaction characteristics between the target traffic object and all reference traffic objects based on the impact factors of all reference traffic objects on the target traffic object.

[0079] In an embodiment of the present invention, by integrating the influence factors of all reference traffic objects on the target traffic object, including the first influence factor, the second influence factor, the third influence factor and the third influence factor mentioned above, the interaction characteristics of the target traffic object and all reference traffic objects can be obtained.

[0080] Exemplarily, in some embodiments of the present invention, step S104 includes the following sub-steps:

[0081] 1. For each reference traffic object, the influencing factor and the attribute vector representing the intrinsic attributes of the reference traffic object are combined into a feature vector.

[0082] In an embodiment of the present invention, considering the impact of the intrinsic attributes of the reference traffic object on the target traffic object, for each reference traffic object, an influencing factor and an attribute vector representing the intrinsic attributes of the reference traffic object are combined into a feature vector. Exemplarily, the intrinsic attributes of the reference traffic object include the vehicle model of the reference traffic object, and the attribute vector is used to represent the vehicle model of the reference traffic object. For example, when an ordinary vehicle encounters a large vehicle or rescue vehicle, it will generally choose to slow down and avoid it. Exemplarily, in an embodiment of the present invention, the attribute vector can be represented using a one-hot encoding.

[0083] For example, in this embodiment of the present invention, for each reference traffic object B j , according to {A i} in the order of impact factors ij , g ij , pt ij , dt ij , then concatenate the arranged influencing factors with the attribute vector to obtain a feature vector, which is x j .

[0084] 2. Extract high-dimensional features of each feature vector.

[0085] For example, in this embodiment of the present invention, for each reference traffic object B j The corresponding eigenvector x j , Deep Neural Networks (DNN) are used to process feature vectors and extract high-dimensional features from them. While retaining important information in the feature vectors, some unnecessary information is filtered out, thereby reducing the computational burden and improving computational efficiency. The embodiments of the present invention do not limit the specific network structure of the deep neural network. For example, it can be a common multi-layer perceptron, etc., as long as it can extract high-dimensional features.

[0086] 3. Aggregate the high-dimensional features of all feature vectors to obtain the interaction features between the target traffic object and all reference traffic objects.

[0087] In this embodiment of the present invention, the high-dimensional features of all feature vectors are aggregated to obtain interaction features between the target traffic object and all reference traffic objects. Aggregation of vectors may include finding the maximum, minimum, average, sum, variance, or median of elements of the same dimension in all high-dimensional features, which is not limited in this embodiment of the present invention.

[0088] For example, in one embodiment of the present invention, all reference traffic objects B j The high-dimensional features are calculated according to {B j} are arranged in order to form a feature matrix. A pooling operation is performed on multiple elements in each dimension of the feature matrix to obtain the interaction features between the target traffic object and all reference traffic objects. It should be noted that the selected pooling operation requires that the result is not affected by the order of the elements.

[0089] For example, referring to traffic object B j The high-dimensional features of the eigenvector are recorded as t j =[t 1j ,t 2j ,…t mj ], m is t j Is the dimension. Then the characteristic matrix T is recorded as:

[0090]

[0091] Wherein, j=1, 2…n.

[0092] Taking the pooling operation as maximum pooling as an example, a pooling operation is performed on multiple elements of each dimension in the feature matrix, that is, the maximum value of all elements in each column of the above feature matrix T is screened out to obtain the interaction features between the target traffic object and all reference traffic objects.

[0093] S105: Predict the behavior of the target traffic object based on the interaction features.

[0094] For example, in some embodiments of the present invention, interaction features, state features of the target traffic object in latent space, and sampled features of map information can be concatenated to generate a concatenated vector. This concatenated vector is then input into a pre-built prediction model for processing to determine the target traffic object's behavior at the next moment. The target traffic object's behavior can be characterized by its displacement relative to the previous moment, and its current speed and direction. The prediction model can be a long short-term memory network model or other temporal recurrent network model, and this embodiment of the present invention is not limited thereto.

[0095] The map information may be information from a pre-built high-precision map, and the sampled features of the map information may be features within a preset range of the target traffic object. For example, a convolutional neural network may be used to encode the map information to obtain map encoding features, and then the map encoding features may be sampled based on the current position of the target traffic object to obtain sampled features of the map information.

[0096] The state characteristics of the target traffic object in latent space can be determined based on the target traffic object's historical trajectory and posterior distribution. The posterior distribution is a probability distribution derived from the distribution of samples (future trajectories of obstacles) and the prior distribution of unknown parameters, using the conditional probability distribution method in probability theory. For example, the posterior distribution is sampled based on the target traffic object's historical position to obtain sampling characteristics of the posterior distribution. The sampling characteristics of the posterior distribution vector and the historical position are then fused to obtain the state characteristics of the target traffic object in latent space.

[0097] The traffic object behavior prediction method provided by the embodiment of the present invention includes: selecting a target traffic object, and taking the target traffic object as the center, selecting a plurality of points within a first preset range as perception points, for reference traffic objects within a second preset range of the target traffic object, taking the reference traffic objects as the field source of a directed distance field, calculating the distance and gradient value of each perception point in the directed distance field, converting the distance and gradient value of the perception point in the directed distance field into an influence factor of the reference traffic object on the behavior of the target traffic object, determining the interaction characteristics of the target traffic object and all reference traffic objects based on the influence factors of all reference traffic objects on the target traffic object, and calculating the interaction characteristics of the target traffic object and all reference traffic objects based on the influence factors of all reference traffic objects on the target traffic object. In order to predict the behavior of a target traffic object based on interaction features, an embodiment of the present invention uses a reference traffic object as the source of a directed distance field, calculates the distance and gradient value of the perception point of the target traffic object in the directed distance field, converts the distance and gradient value into an influence factor of the reference traffic object on the behavior of the target traffic object, and determines the interaction features of the target traffic object and all reference traffic objects based on the influence factors. This can effectively extract information that is unobservable to the outside world, improve prediction accuracy, and the extracted interaction features enable deep neural networks to efficiently characterize and process the interactions of multiple traffic objects, thereby learning highly generalized interaction prediction capabilities.

[0098] The embodiment of the present invention also provides a traffic object behavior prediction device, Figure 2 A schematic diagram of a traffic object behavior prediction device provided by an embodiment of the present invention is shown in FIG. Figure 2 As shown, the traffic object behavior prediction device includes:

[0099] A sensing point determination module 201 is configured to select a target traffic object and, with the target traffic object as the center, select a plurality of points within a first preset range as sensing points;

[0100] A calculation module 202 is configured to calculate the distance and gradient value of each sensing point in the directed distance field with respect to a reference traffic object within a second preset range of the target traffic object, using the reference traffic object as a field source of the directed distance field;

[0101] A conversion module 203 is configured to convert the distance and gradient value of the perception point in the directed distance field into an influence factor of the reference traffic object on the behavior of the target traffic object;

[0102] An interaction feature determination module 204 is configured to determine interaction features between the target traffic object and all reference traffic objects based on the influence factors of all reference traffic objects on the target traffic object;

[0103] The behavior prediction module 205 is configured to predict the behavior of the target traffic object based on the interaction features.

[0104] In some embodiments of the present invention, the perception point determination module 201 includes:

[0105] An envelope box determination submodule, configured to determine the envelope box of the target traffic object;

[0106] The perception point selection submodule is used to select four vertices on the envelope box as perception points.

[0107] In some embodiments of the present invention, the calculation module 202 includes:

[0108] A distance function construction submodule, configured to construct a distance function from a point in the directed distance field to the field source, taking the reference traffic object as the field source of the directed distance field;

[0109] a distance calculation submodule, configured to substitute the coordinates of the perception point into the distance function to calculate the distance of the perception point in the directed distance field;

[0110] The gradient value calculation submodule is used to calculate the gradient value of the distance function at the perception point.

[0111] In some embodiments of the present invention, the conversion module 203 includes:

[0112] a transformation submodule, configured to transform the distance and the gradient value using a preset monotonically decreasing and differentiable transformation function, and transform the distance and the gradient value into a first influencing factor and a second influencing factor;

[0113] a partial derivative calculation submodule, configured to calculate a time partial derivative of the first influencing factor of the sensing point in the global coordinate system as a third influencing factor, while only considering the movement of the reference traffic object in the global coordinate system;

[0114] The total derivative calculation submodule is used to calculate the time total derivative of the first influencing factor of the perception point in the global coordinate system as the fourth influencing factor, taking into account the relative motion of the reference traffic object and the target traffic object in the global coordinate system.

[0115] In some embodiments of the present invention, the transformation submodule includes:

[0116] A first impact factor calculation unit, configured to substitute the distance into the transformation function to obtain a first impact factor;

[0117] a derivation unit, configured to calculate a derivative of the transformation function at the distance;

[0118] The second impact factor calculation unit is used to calculate the product of the derivative and the gradient to obtain the second impact factor.

[0119] In some embodiments of the present invention, the interaction feature determination module 204 includes:

[0120] a vector combining submodule, configured to combine, for each of the reference traffic objects, the influencing factor and an attribute vector representing an intrinsic attribute of the reference traffic object into a feature vector;

[0121] A feature extraction submodule, configured to extract high-dimensional features of each of the feature vectors;

[0122] The feature aggregation submodule is used to aggregate the high-dimensional features of all the feature vectors to obtain the interaction features between the target traffic object and all reference traffic objects.

[0123] In some embodiments of the present invention, the feature aggregation submodule includes:

[0124] A feature combination unit, used for combining the high-dimensional features of all the feature vectors into a feature matrix;

[0125] The pooling unit is used to perform a pooling operation on multiple elements of each dimension in the feature matrix to obtain interaction features between the target traffic object and all reference traffic objects.

[0126] In some embodiments of the present invention, the pooling operation includes maximum pooling, minimum pooling and average pooling.

[0127] The above-mentioned traffic object behavior prediction device can execute the traffic object behavior prediction method provided by any embodiment of the present application, and has the corresponding functional modules and beneficial effects for executing the traffic object behavior prediction method.

[0128] An embodiment of the present application provides an electronic device, Figure 3A schematic diagram of the structure of an electronic device provided for an embodiment of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0129] like Figure 3 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., which is communicatively connected to the at least one processor 11. The memory stores a computer program that can be executed by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. Various programs and data required for the operation of the electronic device 10 can also be stored in the RAM 13. The processor 11, ROM 12, and RAM 13 are connected to each other via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0130] Multiple components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0131] The processor 11 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any other suitable processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the traffic object behavior prediction method.

[0132] In some embodiments, the traffic object behavior prediction method can be implemented as a computer program that is tangibly contained in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the traffic object behavior prediction method described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the traffic object behavior prediction method in any other appropriate manner (e.g., by means of firmware).

[0133] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0134] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0135] In the context of the present invention, computer-readable storage media can be tangible media that can contain or store a computer program for use with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Computer-readable storage media can include but are not limited to electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, computer-readable storage media can be machine-readable signal media. More specific examples of machine-readable storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0136] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0137] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0138] A computing system may include clients and servers. The clients and servers are typically remote from each other and typically interact via a communication network. This client-server relationship arises through computer programs running on the respective computers, creating a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host. This server is a hosting product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosting and VPS services.

[0139] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the traffic object behavior prediction method provided in any embodiment of the present application.

[0140] The computer program product may be implemented by writing computer program code for performing the operations of the present invention in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0141] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.

[0142] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A traffic object behavior prediction method, characterized in that: include: A target traffic object is selected, and a plurality of points within a first preset range with the target traffic object as the center are selected as sensing points; For a reference traffic object within a second preset range of the target traffic object, using the reference traffic object as a field source of a directed distance field, calculating a distance and a gradient value of each of the sensing points in the directed distance field; Converting the distance and gradient value of the perception point in the directed distance field into an influence factor of the reference traffic object on the behavior of the target traffic object; Determining interaction characteristics between the target traffic object and all reference traffic objects based on influence factors of all reference traffic objects on the target traffic object; predicting the behavior of the target traffic object based on the interaction feature; Converting the distance and gradient value of the perception point in the directed distance field into an influence factor of the reference traffic object on the behavior of the target traffic object includes: Transforming the distance and the gradient value using a preset monotonically decreasing and differentiable transformation function to transform the distance and the gradient value into a first influencing factor and a second influencing factor; When only the reference traffic object moves in the global coordinate system, a time partial derivative of the first influencing factor of the sensing point in the global coordinate system is calculated as a third influencing factor; In consideration of the relative motion between the reference traffic object and the target traffic object in the global coordinate system, a time total derivative of the first influencing factor of the perception point in the global coordinate system is calculated as a fourth influencing factor.

2. The traffic object behavior prediction method according to claim 1, characterized in that: Taking the target traffic object as the center, several points within a first preset range are selected as sensing points, including: Determining an envelope box of the target traffic object; Four vertices on the envelope box are selected as sensing points.

3. The traffic object behavior prediction method according to claim 1, characterized in that: Taking the reference traffic object as the field source of the directed distance field, calculating the distance and gradient value of each of the sensing points in the directed distance field includes: Taking the reference traffic object as a field source of a directed distance field, constructing a distance function from a point in the directed distance field to the field source; Substituting the coordinates of the perception point into the distance function, and calculating the distance of the perception point in the directed distance field; Calculate the gradient value of the distance function at the sensing point.

4. The traffic object behavior prediction method according to claim 3, characterized in that: The distance and the gradient value are transformed using a preset monotonically decreasing and differentiable transformation function to transform the distance and the gradient value into a first influencing factor and a second influencing factor, including: Substituting the distance into the transformation function to obtain a first impact factor; calculating a derivative of the transformation function at the distance; The product of the derivative and the gradient is calculated to obtain the second influencing factor.

5. The traffic object behavior prediction method according to any one of claims 1 to 4, characterized in that: Determining interaction characteristics between the target traffic object and all reference traffic objects based on influence factors of all reference traffic objects on the target traffic object includes: For each of the reference traffic objects, combining the influencing factor and an attribute vector representing an intrinsic attribute of the reference traffic object into a feature vector; Extracting high-dimensional features from each of the feature vectors; Aggregate the high-dimensional features of all the feature vectors to obtain interaction features between the target traffic object and all reference traffic objects.

6. The traffic object behavior prediction method according to claim 5, characterized in that: Aggregating the high-dimensional features of all the feature vectors to obtain the interaction features between the target traffic object and all reference traffic objects, including: Combining the high-dimensional features of all the feature vectors into a feature matrix; A pooling operation is performed on multiple elements of each dimension in the feature matrix to obtain interaction features between the target traffic object and all reference traffic objects.

7. The traffic object behavior prediction method according to claim 6, characterized in that: The pooling operation includes maximum pooling, minimum pooling and average pooling.

8. A traffic object behavior prediction device, characterized in that: include: a sensing point determination module, configured to select a target traffic object and, with the target traffic object as the center, select a plurality of points within a first preset range as sensing points; a calculation module, configured to calculate the distance and gradient value of each of the sensing points in the directed distance field with respect to a reference traffic object within a second preset range of the target traffic object, using the reference traffic object as a field source of the directed distance field; a conversion module, configured to convert the distance and gradient value of the perception point in the directed distance field into an influence factor of the reference traffic object on the behavior of the target traffic object; an interaction feature determination module, configured to determine interaction features between the target traffic object and all reference traffic objects based on influence factors of all reference traffic objects on the target traffic object; a behavior prediction module, configured to predict the behavior of the target traffic object based on the interaction features; The conversion module includes: a transformation submodule, configured to transform the distance and the gradient value using a preset monotonically decreasing and differentiable transformation function, and transform the distance and the gradient value into a first influencing factor and a second influencing factor; a partial derivative calculation submodule, configured to calculate a time partial derivative of the first influencing factor of the sensing point in the global coordinate system as a third influencing factor, while only considering the movement of the reference traffic object in the global coordinate system; The total derivative calculation submodule is used to calculate the time total derivative of the first influencing factor of the perception point in the global coordinate system as the fourth influencing factor, taking into account the relative motion of the reference traffic object and the target traffic object in the global coordinate system.

9. An electronic device, characterized in that: include: one or more processors; a memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the traffic object behavior prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the traffic object behavior prediction method according to any one of claims 1 to 7 is implemented.

Images