Construction vehicle trajectory prediction method and device based on physical law constraint

Abstract

The invention provides a construction vehicle trajectory prediction method and device based on physical law constraint. The method comprises the steps that a known track and state information of a target vehicle are acquired, and the state information comprises driving behavior information of the target vehicle and / or action information of the environment on the target vehicle; inputting the known trajectory and the state information into a trained trajectory prediction model, and obtaining a predicted trajectory of the target vehicle output by the trained trajectory prediction model; wherein the trained trajectory prediction model is used for determining and outputting a prediction trajectory according to the known trajectory and the state information. The prediction model is combined with the known trajectory and state information to predict the trajectory of the target vehicle, physical constraints are introduced to accurately simulate the actual behavior of the target vehicle, the precision of construction vehicle trajectory prediction is improved, and it is ensured that the prediction result conforms to the physical law in an actual traffic scene.

Description

Technical Field

[0001] This application belongs to the field of computer technology, and particularly relates to a construction vehicle trajectory prediction method and device based on physical law constraints. Background Art

[0002] Currently, trajectory prediction is used to improve driving safety and solve driving safety problems. Specifically, by accurately predicting the future trajectory of the target vehicle, driving safety can be improved, and risk assessment and decision-making planning of the host vehicle can be realized.

[0003] Trajectory prediction generally uses a construction vehicle trajectory prediction model to predict the trajectory of a construction vehicle based on the trajectory data of the construction vehicle. The model prediction relies on a single data-driven method, resulting in low prediction accuracy. Summary of the Invention

[0004] Embodiments of this application provide a construction vehicle trajectory prediction method, device, electronic device, readable storage medium, and computer program product based on physical law constraints, which can solve the problem of low prediction accuracy.

[0005] In a first aspect, embodiments of this application provide a construction vehicle trajectory prediction method based on physical law constraints, including:

[0006] Obtain the known trajectory and state information of the target vehicle, where the state information includes the driving behavior information of the target vehicle and / or the information of the environment acting on the target vehicle;

[0007] Input the known trajectory and the state information into a trained trajectory prediction model to obtain the predicted trajectory of the target vehicle output by the trained trajectory prediction model;

[0008] Wherein, the trained trajectory prediction model is used to determine and output the predicted trajectory according to the known trajectory and the state information.

[0009] In a second aspect, embodiments of this application provide a construction vehicle trajectory prediction device, including:

[0010] An acquisition module, configured to obtain the known trajectory and state information of the target vehicle, where the state information includes the driving behavior information of the target vehicle and / or the information of the environment acting on the target vehicle;

[0011] A prediction module, configured to input the known trajectory and the state information into a trained trajectory prediction model to obtain the predicted trajectory of the target vehicle output by the trained trajectory prediction model;

[0012] Wherein, the trained trajectory prediction model is used to determine and output the predicted trajectory according to the known trajectory and the state information.

[0013] In a third aspect, an embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method described in any one of the above first aspects is implemented.

[0014] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the method described in any one of the above first aspects is implemented.

[0015] In a fifth aspect, an embodiment of the present application provides a computer program product, and when the computer program product runs on an electronic device, the electronic device is enabled to execute the method described in any one of the above first aspects.

[0016] The beneficial effects of the embodiments of the present application compared with the prior art are as follows:

[0017] In the embodiments of the present application, by obtaining the known trajectory and status information of the target vehicle, where the status information includes the driving behavior information of the target vehicle and / or the information of the environment acting on the target vehicle; inputting the known trajectory and status information into the trained trajectory prediction model to obtain the predicted trajectory of the target vehicle output by the trained trajectory prediction model; wherein, the trained trajectory prediction model is used to determine and output the predicted trajectory according to the known trajectory and status information, so that the prediction model combines the known trajectory and status information to perform trajectory prediction on the target vehicle, introduces physical constraints to accurately simulate the actual behavior of the target vehicle, improves the accuracy of the trajectory prediction of the construction vehicle, and ensures that the prediction result conforms to the physical laws in the actual traffic scenario.

[0018] It can be understood that the beneficial effects of the above second aspect to fifth aspect can refer to the relevant descriptions in the above first aspect, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 is the first flowchart of the construction vehicle trajectory prediction method provided by an embodiment of the present application;

[0021] Figure 2 is the structural diagram of the trained trajectory prediction model provided by an embodiment of the present application;

[0022] Figure 3 It is the second process schematic diagram of the construction vehicle trajectory prediction method provided by an embodiment of the present application;

[0023] Figure 4 It is the structural schematic diagram of the construction vehicle trajectory prediction device provided by an embodiment of the present application;

[0024] Figure 5 It is the structural schematic diagram of the electronic device provided by an embodiment of the present application. Detailed implementation manners

[0025] In the following description, for the purpose of illustration rather than limitation, specific details such as specific system architectures, technologies, etc. are presented to thoroughly understand the embodiments of the present application. However, those skilled in the art should clearly understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present application.

[0026] It should be understood that when used in the specification and appended claims of the present application, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0027] It should also be understood that the term "and / or" as used in the specification and appended claims of the present application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0028] As used in the specification and appended claims of the present application, the term "if" can be interpreted as "when", "once", "in response to determining", or "in response to detecting" according to the context. Similarly, the phrase "if determined" or "if [the described condition or event] is detected" can be interpreted as meaning "once determined", "in response to determining", "once [the described condition or event] is detected", or "in response to detecting [the described condition or event]" according to the context.

[0029] In addition, in the description of the specification and appended claims of the present application, the terms "first", "second", "third", etc. are only used for differentiating descriptions and cannot be understood as indicating or implying relative importance.

[0030] References to "one embodiment" or "some embodiments" or the like described in the specification of this application mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in one or more embodiments of this application. Thus, statements such as "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments", etc. that appear in different places in this specification do not necessarily all refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized. The terms "comprising", "including", "having" and their variants all mean "including but not limited to", unless otherwise specifically emphasized.

[0031] In one embodiment, Figure 1 is the first schematic flow diagram of the construction vehicle trajectory prediction method provided by an embodiment of this application. As Figure 1 shown, the method includes:

[0032] S11: Obtain the known trajectory and status information of the target vehicle.

[0033] Among them, the status information includes the driving behavior information of the target vehicle and / or the information of the environment acting on the target vehicle. The selection of the status information needs to be determined after in-depth analysis of the physical and behavioral characteristics of the construction vehicle. The status information is the information that affects the movement of the target vehicle. The driving behavior information may include information in situations such as moving forward to the desired position in an optimal manner in an actual traffic scenario, safe following behavior, etc., and the information of the environment acting on the target vehicle may include information in situations such as collision avoidance caused by obstacles and lane keeping requirements caused by lanes in an actual traffic scenario. The status information characterizes the physical laws in the actual traffic scenario.

[0034] The known trajectory is sequence data, including historical position points for a period of time.

[0035] S12: Input the known trajectory and status information into the trained trajectory prediction model to obtain the predicted trajectory of the target vehicle output by the trained trajectory prediction model.

[0036] Among them, the trained trajectory prediction model is used to determine and output the predicted trajectory according to the known trajectory and status information.

[0037] Among them, the expression of the input data of the model is X = [V track , V physics , V track is the known trajectory, and V physics is the status information.

[0038] In an application, the trained trajectory prediction model processes historical position points to learn the past motion patterns of the target vehicle; the trained trajectory prediction model processes state information to learn the physical laws in the actual traffic scenario. The future trajectory of the target vehicle is predicted by learning the past motion patterns and physical laws.

[0039] It can be understood that the existing prediction model uses a data-driven approach to predict the trajectory of construction vehicles. This prediction method lacks a physical interpretation of the internal mechanism, resulting in distrust of the decision-making results. In this embodiment, the trained trajectory prediction model processes state information to learn the physical laws in the actual traffic scenario, provides the physical meaning of the data, enhances the interpretability of the model, and improves the trust in the decision-making results.

[0040] The model of this embodiment combines the data-driven approach and the physical-driven approach, and can accurately predict the trajectory of the target vehicle in a traffic scenario with complex and changeable environments, reducing the over-reliance of the model on the sample size and feature generalization in the data-driven approach.

[0041] Moreover, the known trajectory is sequence data, including historical position points for a period of time, enabling the model to learn the past motion patterns of the target vehicle and improve the performance of the model.

[0042] In this embodiment, the known trajectory and state information of the target vehicle are obtained. The state information includes the driving behavior information of the target vehicle and / or the information of the influence of the environment on the target vehicle; the known trajectory and state information are input into the trained trajectory prediction model to obtain the predicted trajectory of the target vehicle output by the trained trajectory prediction model; wherein, the trained trajectory prediction model is used to determine and output the predicted trajectory according to the known trajectory and state information, so that the prediction model combines the known trajectory and state information to predict the trajectory of the target vehicle, introduces physical constraints to accurately simulate the actual behavior of the target vehicle, improves the accuracy of the construction vehicle trajectory prediction, and ensures that the prediction results conform to the physical laws in the actual traffic scenario.

[0043] In one embodiment, Figure 2 is a schematic structural diagram of the trained trajectory prediction model provided by an embodiment of the present application. As Figure 2 shown, the trained trajectory prediction model includes a trained trajectory encoding module, a trained social force rule module, a trained cascade module, and a trained trajectory decoding module.

[0044] In a possible implementation manner, the trained trajectory prediction model can be SF-GRU. SF-GRU is based on the GRU (Gated Recurrent Unit) structure and is constructed by combining social force rules. Among them, the trajectory encoding module is a GRU encoder. The trajectory decoding module is a GRU decoder.

[0045] Step S12 includes:

[0046] S121: Input the known trajectory into the trained trajectory encoding module, perform encoding processing on the known trajectory through the trained trajectory encoding module to obtain the final hidden state; and input the final hidden state into the trained cascade module;

[0047] where the expression of the known trajectory is V track =[X i,t-T+1 ,…,X i,j ,…,X i,t , the known trajectory includes T position coordinate points, X i,j =(x i,j ,y i,j ) is the x and y components in the coordinates of the target vehicle i at time j, and T is the memory time range of the known trajectory. V track ∈R C×M×T , C is the number of features, the value is determined to be 2 according to the x and y coordinates, M is the number of trajectories, and T is the length of the trajectory. The coordinates can be Cartesian coordinates or other coordinates.

[0048] In a possible implementation, the reset gate in the trained GRU encoder is responsible for determining the amount of previous state information to be forgotten. The formula for the reset gate: r t =σ(W r V track +U r h t-1 ). The update gate in the trained GRU encoder is responsible for determining the amount of new state information to be introduced into the current state. The formula for the update gate: z t =σ(W z V track +U z h t-1 ). Where σ is the sigmoid activation function, and the value output by σ ranges from [0,1], realizing fine control of information flow. Based on the reset gate and the update gate, the formula for calculating the hidden state at time t (1≤t≤T) is h t =(1 - z t )h t-1 +z t h~ t , h~ t =tanh(WV track +U(r t ⊙h t-1 ))), h t-1 is the previously stored content, h t is the candidate storage content, and W and U are the weight matrices learned by the model. After calculating the sequence data, the final hidden state is obtained, and the final hidden state represents the information of the entire sequence data.

[0049] Use the GRU encoder to process the data, fully explore the potential information and features of the known trajectory of the target vehicle, accurately process the known trajectory, effectively solve problems such as gradient messages, and improve the stability and accuracy of the model when processing time-series data.

[0050] S122: Input the status information into the trained social force rule module, and determine the social force acting on the target vehicle according to the status information through the trained social force rule module; and input the social force into the trained cascade module.

[0051] Among them, the social force rule represents the physical rules in the actual traffic scenario. Determine the social force rule according to the actual traffic scenario. The social force rule describes the physical meaning implied by the status information and the interaction behavior of the target vehicle (physical interaction between construction vehicles and the interaction between construction vehicles and the environment). Specifically, it may include the driving force rule for the expected position at the next time, the attraction rule, repulsion rule, and boundary force rule of the surrounding environment, etc.

[0052] In the application, the model uses the trained social force rule module to supplement and enhance the status information, better simulate the actual situation of the target vehicle on the road, and more accurately and detailedly describe the social force received by the target vehicle on the road.

[0053] In a possible implementation, in the actual traffic scenario, the target vehicle will move forward to the expected position in an optimal manner while maintaining driving safety. The attraction rule for the expected position at the next time is that in the absence of external interference, the target vehicle will move forward along the expected path at a set speed. During this process, the target vehicle will experience a certain response time, and through acceleration and deceleration adjustment, achieve a smooth transition from the current speed to the expected speed, and the acceleration of the target vehicle does not exceed the maximum capacity limit. Under the rule, the target vehicle will be subject to a driving force, and the driving force is used to drive the target vehicle towards the expected position. Correspondingly, the social force includes the driving force.

[0054] Correspondingly, the driving behavior information includes the maximum acceleration, expected speed, direction of the expected speed, first speed, and response time of the target vehicle. The response time is the time required for the first speed to change to the expected speed.

[0055] Step S121 includes:

[0056] S21: Input the maximum acceleration, expected speed, direction of the expected speed, first speed, and response time into the trained social force rule module, and determine the driving force acting on the target vehicle according to the maximum acceleration, expected speed, direction of the expected speed, first speed, and response time through the trained social force rule module.

[0057] In the application, the trained social force rule module uses the formula Calculate the driving force.

[0058] Where, i refers to the target vehicle, is the driving force, a 1 is the first coefficient of the force, and the first coefficient is set according to the scenario. is the maximum acceleration, is the desired speed at time t, is the direction of the desired speed at time t, v i (t) is the first speed (actual speed) at time t, τ i is the response time.

[0059] In a possible implementation, in an actual traffic scenario, when the target vehicle needs to follow the vehicle in front, it will safely follow the vehicle in front and exhibit a safe following behavior. The attraction rule of the surrounding environment is that the target vehicle follows the vehicle in front at a safe distance and a safe speed. Under this rule, the target vehicle will be subject to an attraction force, which is used to drive the target vehicle to follow the vehicle in front, and the social force includes the attraction force.

[0060] Correspondingly, the driving behavior information also includes the first maximum deceleration of the target vehicle, the reaction time of the driver, the attraction direction, the second speed of the vehicle in front, the second maximum deceleration of the vehicle in front, and the safe distance between the target vehicle and the vehicle in front. The attraction direction is determined according to the driving direction of the vehicle in front, and the vehicle in front is the vehicle that the target vehicle follows.

[0061] In a possible implementation, the attraction direction can be the same as the driving direction of the vehicle in front, which can describe the situation where the target vehicle is affected by the vehicle in front.

[0062] For example, when the vehicle in front goes straight, the attraction direction is the same as the direction in which the vehicle in front goes straight. When the vehicle in front turns right, the attraction direction is the same as the direction in which the vehicle in front turns right. When the vehicle in front is about to turn right, there are two driving directions. Before turning right, the target vehicle follows the current driving direction of the vehicle in front, and the attraction direction is the same as the current driving direction of the vehicle in front. When turning right, the target vehicle follows the vehicle in front to turn right, and the attraction direction is the same as the driving direction of the vehicle in front after turning right.

[0063] Step S121 includes:

[0064] S22: Input the first maximum deceleration, the reaction time of the driver, the first speed, the attraction direction, the second speed, the second maximum deceleration, and the safe distance into the trained social force rule module. According to the trained social force rule, determine the safe speed for the target vehicle to maintain a distance from the vehicle in front based on the first maximum deceleration, the first speed, the second speed, the second maximum deceleration, and the safe distance; determine the attraction force according to the safe speed, the reaction time, the maximum acceleration, and the attraction direction.

[0065] In the application, in the trained social force rule module, the formula Calculate the safe speed. Among them, the vehicle ahead is i + 1, is the safe speed, g i (t) is the safe distance at time t, v i+1 (t) is the second speed (actual speed) at time t, b i is the first maximum deceleration, b i+1 is the second maximum deceleration, δ i is the reaction time. The trained social force rule module uses the formula to calculate the attraction force, where is the attraction force, a 2 is the second coefficient of the force, and the second coefficient is set according to the scenario, n i,i+1 is the attraction direction.

[0066] In a possible implementation, in an actual traffic scenario, when the target vehicle encounters an obstacle, it will avoid the obstacle to prevent collision. The repulsive force rule of the surrounding environment is for the target vehicle to stay away from the surrounding obstacles and maintain a safe distance from the obstacles. Under this rule, the target vehicle will be subject to a repulsive force, and the repulsive force is used to drive the target vehicle away from the obstacle. The social force includes the repulsive force.

[0067] Among them, the obstacles include surrounding pedestrians, other construction vehicles, static obstacles, etc. Other construction vehicles can be the vehicles behind in the same lane or adjacent lanes. There are multiple obstacles around the target vehicle, and multiple sets of action information corresponding to the multiple obstacles are obtained.

[0068] Correspondingly, the action information includes the repulsive force intensity, the first distance, the repulsive force action range, and the first unit vector between the target vehicle and the obstacles in the environment. The first unit vector is the unit vector pointing from the center of the obstacle to the center of the target vehicle.

[0069] Among them, the repulsive force action range controls the attenuation speed of the repulsive force with distance and defines the effective action range of the repulsive force. Generally, the larger the action range, the farther the repulsive force can act; the smaller the action range, the more effective the repulsive force is at close range. When the first distance is within the repulsive force action range, the repulsive force is obvious. When the first distance approaches or exceeds the repulsive force action range, the repulsive force decreases and approaches zero. The repulsive force action range can be determined according to factors such as the type and size of the construction vehicle, the driving speed, and the road environment.

[0070] The repulsive force action range can be specifically determined by analyzing the historical trajectory data of the construction vehicle. For example, by statistically analyzing the interactions of construction vehicles in different situations, a reasonable value can be fitted. A reasonable value can also be predicted through a model.

[0071] The repulsive force intensity can be determined through a physically-driven theoretical framework and data-driven experimental optimization. For example, based on the theoretical basis of the model's formula, real data is used for parameter fitting and optimization, or experiments and sensitivity analysis are conducted to verify the rationality of the parameters to determine reasonable values.

[0072] Step S122 includes:

[0073] S23: Input the repulsive force intensity, the first distance, the repulsive force range of action, and the first unit vector into the trained social force rule module. The trained social force rule module determines the repulsive force acting on the target vehicle according to the repulsive force intensity, the first distance, the repulsive force range of action, and the first unit vector.

[0074] In the application, the trained social force rule module uses the formula to calculate the repulsive force. is the repulsive force between the target vehicle i and the obstacle j, is the repulsive force intensity, is the first distance, is the repulsive force range of action, is the first unit vector.

[0075] In a possible implementation, in an actual traffic scenario, when the target vehicle is driving on the lane, it generally drives near the center line of the lane and maintains a certain distance from the roadside. The boundary force rule for the surrounding environment is the lane-keeping requirement, driving near the center line of the lane and maintaining a certain distance from the roadside. Under this rule, the target vehicle is subject to the boundary force, and the boundary force is used to drive the target vehicle to be located on the center line of the lane. The social force includes the boundary force.

[0076] Correspondingly, the action information includes the boundary force intensity between the lane boundary and the target vehicle, the second distance, the boundary force range of action, and the second unit vector. The second unit vector is the unit vector pointing from the lane boundary to the center of the target vehicle.

[0077] Among them, the second distance can be the vertical distance from the center of the target vehicle to the lane boundary. For example, if the target vehicle is driving on the center line of the lane and the center of the construction vehicle coincides with the center line of the lane, the second distance is the vertical distance from the center of the target vehicle to any lane boundary. If the target vehicle is driving on an irregular lane (such as a curve, a narrow section, etc.), the vertical distances from the target vehicle to the two boundaries are different and not constant values. The vertical distance is determined according to the position of the target vehicle and the selected boundary.

[0078] The boundary force intensity can be determined through a physically-driven theoretical framework and data-driven experimental optimization. For example, based on the theoretical basis of the model's formula, real data is used for parameter fitting and optimization, or experiments and sensitivity analysis are conducted to verify the rationality of the parameters to determine reasonable values.

[0079] The boundary force action range can be adjusted through data-driven optimization, experimental analysis, and by combining the requirements of specific scenarios, and then through model optimization, so as to balance the boundary force action range between accuracy and applicability and improve the actual effectiveness of the model.

[0080] Step S122 includes:

[0081] Step S24: Input the boundary force intensity, the second distance, the boundary force action range, and the second unit vector into the trained social force rule module, and determine the boundary force acting on the target vehicle through the trained social force rule module according to the boundary force intensity, the second distance, the boundary force action range, and the second unit vector.

[0082] In the application, the trained social force rule module uses the formula to calculate the boundary force. is the boundary force, A ib is the boundary force intensity, d ib is the second distance, B ib is the boundary force action range, n ib is the second unit vector.

[0083] It can be understood that during the driving process of the target vehicle, due to actual situations, it is affected by at least one of the above social forces, and the corresponding trained social force rule module will output at least one social force.

[0084] S123: After the final hidden state and the social force are input into the trained cascade module, the trained cascade module performs matrix concatenation processing on the final hidden state and the social force to obtain concatenated data, and inputs the concatenated data into the trained trajectory decoding module;

[0085] In the application, the trained cascade module performs matrix concatenation processing on the final hidden state and the social force through the formula O = W tracks O tracks + W SF O SF + b, where W tracks and W SF are learnable parameters, O tracks is the final hidden state, O SF is the social force, and b is the bias coefficient. O tracks and O SF are data in matrix form.

[0086] S124: After the concatenated data is input into the trained trajectory decoding module, the trained trajectory decoding module decodes the concatenated data to generate and output the predicted trajectory.

[0087] In a possible implementation, the trained GRU decoder uses the formula to generate a predicted trajectory. Among them, the Relu function is the activation function.

[0088] Specifically, the predicted trajectory The expression is Among them, is the predicted position at time j, and H is the prediction time range of the predicted trajectory.

[0089] In this embodiment, the state information is supplemented and enhanced by the social force rule, which can handle various social forces experienced by the target vehicle, enabling the model to more accurately simulate the actual behavior of the target vehicle in the actual and complex traffic environment, improving the accuracy of model prediction, obtaining more accurate prediction results, and making the prediction results of the model more in line with the physical laws in the actual traffic scenario; and through the social force rule, the model can intuitively and accurately consider the interactions in the actual traffic scenario, accurately describe the changes of relevant variables in the prediction process, enhance the interpretability of the model, be able to provide the reasons for the prediction, make the prediction results of the model easier to understand and accept, so as to improve the safety of autonomous driving and the operation efficiency of traffic flow.

[0090] Moreover, by introducing the social force rule to predict the trajectory, the real-time problem caused by the overly complex deep learning architecture can be reduced, so as to balance the real-time performance and complexity of the model.

[0091] In one embodiment, Figure 3 is the second process schematic diagram of the construction vehicle trajectory prediction method provided by an embodiment of the present application. As Figure 3 shown, before obtaining the known trajectory and state information of the target vehicle, it further includes:

[0092] S31: Obtain training samples and labels.

[0093] Among them, the training samples include the trajectory point samples and state information samples of each construction vehicle sample, and the labels include the real position points of each construction vehicle sample.

[0094] S32: Use the training samples to train the trajectory prediction model until the value of the model loss function is less than the preset loss value, and obtain the trained trajectory prediction model.

[0095] Among them, the model loss function is used to calculate the difference value between the output result of the trajectory prediction model and the label.

[0096] In application, the model loss function can be the root mean square error function. Since the movement of the construction vehicle includes horizontal and vertical dimensions, the corresponding model loss function calculates the horizontal and vertical losses.

[0097] The formula of the model loss function is Among them, is the output result of the model, x t+H , y t+H is the real position point, and N is the total number of trajectory point samples in the training samples.

[0098] In this embodiment, by obtaining training samples and labels, the training samples include trajectory point samples and state information samples of each construction vehicle sample, and the labels include the real position points of each construction vehicle sample. Using the training samples, train the trajectory prediction model until the value of the model loss function is less than the preset loss value, and obtain the trained trajectory prediction model, and obtain a model that can perform prediction tasks and accurately predict the trajectory of construction vehicles.

[0099] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The order of execution of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application. And the data collection in the above embodiments is compliant, and its use or implementation does not involve harming the public interest.

[0100] Corresponding to the method described in the above embodiments, for the sake of convenience of description, only the parts related to the embodiments of the present application are shown.

[0101] In one embodiment, Figure 4 is a schematic structural diagram of a construction vehicle trajectory prediction device provided by an embodiment of the present application. As Figure 4 shown, a construction vehicle trajectory prediction device includes:

[0102] An acquisition module 10, configured to acquire the known trajectory and state information of the target vehicle, and the state information includes the driving behavior information of the target vehicle and / or the information of the environment acting on the target vehicle.

[0103] A prediction module 11, configured to input the known trajectory and state information into the trained trajectory prediction model, and obtain the predicted trajectory of the target vehicle output by the trained trajectory prediction model;

[0104] Among them, the trained trajectory prediction model is used to determine and output the predicted trajectory according to the known trajectory and state information.

[0105] In one embodiment, the prediction module is specifically configured to input the known trajectory into the trained trajectory encoding module, perform encoding processing on the known trajectory through the trained trajectory encoding module to obtain the final hidden state; and input the final hidden state into the trained cascade module; input the state information into the trained social force rule module, determine the social force acting on the target vehicle according to the state information through the trained social force rule module; and input the social force into the trained cascade module, where the social force rule represents the physical rules in the actual traffic scenario; after the final hidden state and the social force are input into the trained cascade module, perform matrix concatenation processing on the final hidden state and the social force through the trained cascade module to obtain the concatenated data, and input the concatenated data into the trained trajectory decoding module; after the concatenated data is input into the trained trajectory decoding module, decode the concatenated data through the trained trajectory decoding module to generate and output the predicted trajectory. The trained trajectory prediction model includes the trained trajectory encoding module, the trained social force rule module, the trained cascade module, and the trained trajectory decoding module.

[0106] In one embodiment, the prediction module is specifically configured to input the maximum acceleration, the desired speed, the direction of the desired speed, the first speed, and the response time into the trained social force rule module, and determine the driving force acting on the target vehicle according to the maximum acceleration, the desired speed, the direction of the desired speed, the first speed, and the response time through the trained social force rule module. The driving force is used to drive the target vehicle towards the desired position, and the social force includes the driving force.

[0107] In one embodiment, the prediction module is specifically configured to input the first maximum deceleration, the driver's reaction time, the first speed shown, the attraction direction, the second speed, the second maximum deceleration, and the safety distance into the trained social force rule module, and determine the safe speed for the target vehicle to maintain a distance from the vehicle in front according to the first maximum deceleration, the first speed shown, the second speed, the second maximum deceleration, and the safety distance through the trained social force rule; determine the attraction force according to the safe speed, the reaction time, the maximum acceleration, and the attraction direction. The attraction force is used to drive the target vehicle to follow the vehicle in front, and the social force includes the attraction force.

[0108] In one embodiment, the prediction module is specifically configured to input the repulsive force intensity, the first distance, the repulsive force action range, and the first unit vector into the trained social force rule module, and determine the repulsive force acting on the target vehicle according to the repulsive force intensity, the first distance, the repulsive force action range, and the first unit vector through the trained social force rule module. The repulsive force is used to drive the target vehicle away from the obstacle, and the social force includes the repulsive force.

[0109] In one embodiment, the prediction module is specifically configured to input the boundary force intensity, the second distance, the boundary force action range, and the second unit vector into the trained social force rule module. The trained social force rule module determines the boundary force acting on the target vehicle according to the boundary force intensity, the second distance, the boundary force action range, and the second unit vector. The boundary force is used to drive the target vehicle to be located on the center line of the lane, and the social force includes the boundary force.

[0110] In one embodiment, the device further includes a training module.

[0111] The training module is configured to obtain training samples and labels. The training samples include trajectory point samples and status information samples of each construction vehicle sample, and the labels include the true position points of each construction vehicle sample. The training samples are used to train the trajectory prediction model until the value of the model loss function is less than a preset loss value, and a trained trajectory prediction model is obtained. The model loss function is used to calculate the difference value between the output result of the trajectory prediction model and the label.

[0112] Figure 5 The figure is a schematic structural diagram of an electronic device provided in an embodiment of the present application. As Figure 5 shown, the electronic device 2 in this embodiment includes at least one processor 20 ( Figure 5 only one is shown in the figure), a memory 21, and a computer program 22 stored in the memory 21 and executable on the at least one processor 20. When the processor 20 executes the computer program 22, the steps in any of the above method embodiments are implemented.

[0113] The electronic device 2 may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The electronic device 2 may include, but is not limited to, a processor 20 and a memory 21. Those skilled in the art can understand that Figure 5 this is only an example of the electronic device 2, and does not constitute a limitation on the electronic device 2. It may include more or fewer components than shown in the figure, or combine some components, or different components. For example, it may further include input / output devices, network access devices, etc.

[0114] The processor 20 may be a Central Processing Unit (CPU), and the processor 20 may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0115] In some embodiments, the memory 21 may be an internal storage unit of the electronic device 2, such as the hard disk or memory of the electronic device 2. In other embodiments, the memory 21 may also be an external storage device of the electronic device 2, such as a plug-in hard disk equipped on the electronic device 2, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. Further, the memory 21 may also include both the internal storage unit and the external storage device of the electronic device 2. The memory 21 is used to store an operating system, application programs, a BootLoader, data, and other programs, such as the program code of the computer program, etc. The memory 21 may also be used to temporarily store data that has been output or will be output.

[0116] It should be noted that for the content such as information interaction and execution process between the above-mentioned devices / units, since it is based on the same concept as the method embodiment of the present application, for its specific functions and the technical effects brought, reference may be specifically made to the method embodiment part, and details are not described herein again.

[0117] Those skilled in the art can clearly understand that, for the convenience and conciseness of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. In addition, the specific names of the functional units and modules are only for the convenience of mutual distinction and do not limit the protection scope of this application. The specific working processes of the units and modules in the above system can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated here.

[0118] An embodiment of this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the foregoing method embodiments can be implemented.

[0119] An embodiment of this application provides a computer program product. When the computer program product runs on an electronic device, the electronic device can be made to execute the steps in the foregoing method embodiments.

[0120] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-mentioned method embodiments of this application, a computer program can be used to instruct the relevant hardware to complete. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps in the foregoing method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium can at least include: any entity or device that can carry the computer program code to the photographing device / terminal device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disc, etc. In some cases, the computer-readable medium cannot be an electrical carrier signal and a telecommunication signal.

[0121] In the above embodiments, the descriptions of the various embodiments each have their own emphasis. For parts not detailed or recorded in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0122] Those of ordinary skill in the art will appreciate that the units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of this application.

[0123] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network device and method can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical or other forms.

[0124] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0125] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. And these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included in the protection scope of this application.

Claims

1. A construction vehicle trajectory prediction method based on physical law constraints, characterized in that: include: Acquire known trajectory and state information of the target vehicle, wherein the state information includes driving behavior information of the target vehicle and / or information on the effect of the environment on the target vehicle; Inputting the known trajectory and the state information into a trained trajectory prediction model to obtain a predicted trajectory of the target vehicle output by the trained trajectory prediction model; The trained trajectory prediction model is used to determine and output the predicted trajectory based on the known trajectory and the state information.

2. The method according to claim 1, characterized in that The trained trajectory prediction model includes a trained trajectory encoding module, a trained social force rule module, a trained cascade module and a trained trajectory decoding module; Inputting the known trajectory and the state information into a trained trajectory prediction model to obtain a predicted trajectory of the target vehicle output by the trained trajectory prediction model includes: Inputting the known trajectory into the trained trajectory encoding module, encoding the known trajectory through the trained trajectory encoding module to obtain a final hidden state; and inputting the final hidden state into the trained cascade module; Inputting the state information into the trained social force rule module, determining the social force acting on the target vehicle according to the state information by the trained social force rule module; and inputting the social force into the trained cascade module, wherein the social force rule represents the physical rule in the actual traffic scene; After the final hidden state and the social force are input into the trained cascade module, the final hidden state and the social force are subjected to matrix cascade processing by the trained cascade module to obtain cascade data, and the cascade data is input into the trained trajectory decoding module; After the cascade data is input into the trained trajectory decoding module, the cascade data is decoded by the trained trajectory decoding module to generate and output the predicted trajectory.

3. The method according to claim 2, characterized in that The driving behavior information includes the maximum acceleration, the desired speed, the direction of the desired speed, the first speed, and the response time of the target vehicle, wherein the response time is the time required for the first speed to change to the desired speed; The step of inputting the state information into the trained social force rule module, and determining the social force acting on the target vehicle according to the state information by the trained social force rule module, comprises: The maximum acceleration, the expected speed, the direction of the expected speed, the first speed, and the response time are input into the trained social force rule module, and the trained social force rule module determines the driving force acting on the target vehicle according to the maximum acceleration, the expected speed, the direction of the expected speed, the first speed, and the response time. The driving force is used to drive the target vehicle toward the expected position, and the social force includes the driving force.

4. The method according to claim 3, characterized in that The driving behavior information also includes a first maximum deceleration of the target vehicle, a driver's reaction time, an attraction direction, a second speed of a preceding vehicle, a second maximum deceleration of the preceding vehicle, and a safety distance between the target vehicle and the preceding vehicle, wherein the attraction direction is determined according to the driving direction of the preceding vehicle, and the preceding vehicle is a vehicle followed by the target vehicle; The step of inputting the state information into the trained social force rule module, and determining the social force acting on the target vehicle according to the state information by the trained social force rule module, comprises: The first maximum deceleration, the driver's reaction time, the displayed first speed, the attraction direction, the second speed, the second maximum deceleration, and the safety distance are input into the trained social force rule module, and a safe speed at which the target vehicle maintains a distance from the leading vehicle is determined according to the trained social force rule based on the first maximum deceleration, the first speed, the second speed, the second maximum deceleration, and the safety distance; an attraction force is determined according to the safe speed, the reaction time, the maximum acceleration, and the attraction direction, and the attraction force is used to drive the target vehicle to follow the leading vehicle, and the social force includes the attraction force.

5. The method according to claim 4, characterized in that The action information includes the repulsive force strength between the target vehicle and the obstacle in the environment, the first distance, the repulsive force action range, and the first unit vector, where the first unit vector is a unit vector from the center of the obstacle to the center of the target vehicle; The step of inputting the state information into the trained social force rule module, and determining the social force acting on the target vehicle according to the state information by the trained social force rule module, comprises: The repulsive force strength, the first distance, the repulsive force range, and the first unit vector are input into the trained social force rule module, and the trained social force rule module determines the repulsive force acting on the target vehicle according to the repulsive force strength, the first distance, the repulsive force range, and the first unit vector. The repulsive force is used to drive the target vehicle away from the obstacle, and the social force includes the repulsive force.

6. The method according to claim 5, characterized in that The action information includes the boundary force strength between the lane boundary and the target vehicle, a second distance, a boundary force action range, and a second unit vector, where the second unit vector is a unit vector from the lane boundary to the center of the target vehicle; The step of inputting the state information into the trained social force rule module, and determining the social force acting on the target vehicle according to the state information by the trained social force rule module, comprises: The boundary force strength, the second distance, the boundary force range, and the second unit vector are input into the trained social force rule module, and the trained social force rule module determines the boundary force acting on the target vehicle according to the boundary force strength, the second distance, the boundary force range, and the second unit vector. The boundary force is used to drive the target vehicle to be located on the center line of the lane, and the social force includes the boundary force.

7. The method according to any one of claims 1 to 6, characterized in that: Before obtaining the known track and state information of the target vehicle, the method further includes: Acquire training samples and labels, wherein the training samples include trajectory point samples and state information samples of each construction vehicle sample, and the labels include real position points of each construction vehicle sample; Using the training samples, training the trajectory prediction model until the value of the model loss function is less than a preset loss value, thereby obtaining a trained trajectory prediction model; The model loss function is used to calculate the difference between the output result of the trajectory prediction model and the label.

8. A construction vehicle trajectory prediction device, characterized in that: include: An acquisition module, used to acquire known trajectory and state information of a target vehicle, wherein the state information includes driving behavior information of the target vehicle and / or information on the effect of the environment on the target vehicle; A prediction module, used for inputting the known trajectory and the state information into a trained trajectory prediction model to obtain a predicted trajectory of the target vehicle output by the trained trajectory prediction model; The trained trajectory prediction model is used to determine and output the predicted trajectory based on the known trajectory and the state information.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images