Methods, devices, storage media and electronic equipment for constructing dynamic models

Abstract

This specification discloses a method, apparatus, storage medium, and electronic device for constructing dynamic models of vehicles. During model construction, actual state data generated by the vehicle and simulated state data generated by the dynamic model can be determined based on historical control data of the vehicle. Then, based on the state error between the actual and simulated state data, a feedforward neural network is used to predict the compensation values ​​of each model parameter in the dynamic model. Finally, based on the compensation values ​​of each model parameter, the parameters in the dynamic model are adjusted. By predicting the compensation values ​​of model parameters through a feedforward neural network and then adjusting the parameters of the dynamic model, automated adjustment of the dynamic model parameters is achieved, shortening the fitting time of the dynamic model and improving the fitting efficiency.

Description

Technical Field

[0001] This application relates to the field of simulation and testing technology, and in particular to methods, devices, storage media and electronic devices for constructing dynamic models. Background Technology

[0002] The actual application scenarios of unmanned equipment are complex and varied. In order to ensure the safe operation of unmanned equipment, its various performance characteristics need to be tested before it is put into use.

[0003] Taking unmanned vehicles as an example, due to the high cost of real-vehicle testing, simulation testing is currently the most common method. This involves constructing a dynamic model of the unmanned vehicle and testing and verifying various algorithms in a simulation environment. Whether the dynamic model can accurately reflect the motion characteristics of the unmanned vehicle has a significant impact on the accuracy of the simulation test.

[0004] Existing technologies for building vehicle dynamics models for autonomous vehicles first collect real-world vehicle data under different driving conditions. This real-world data includes control data (throttle opening, steering wheel position, etc.) and vehicle state data (actual speed, actual acceleration, driving direction, etc.). Then, the control data from this real-world data is input into the vehicle dynamics model to obtain the simulation state data.

[0005] Finally, based on human experience, the model parameters in the vehicle dynamics model are manually adjusted by comparing the differences between real vehicle state data and simulation state data. These model parameters include, at a minimum, throttle opening in the powertrain, transmission ratio in the transmission system, and steering ratio in the steering system.

[0006] However, manually adjusting model parameters often requires a lot of debugging, has a long adjustment cycle, low parameter tuning efficiency, and high labor costs. Summary of the Invention

[0007] This specification provides a method, apparatus, storage medium, and electronic device for constructing a dynamic model, which partially solves the problems in the prior art.

[0008] The embodiments in this specification adopt the following technical solutions:

[0009] This specification provides a method for constructing a dynamic model, including:

[0010] Acquire historical control data and actual status data of the vehicle during operation;

[0011] The control data is input into the dynamic model of the vehicle to determine the simulation state data generated by the dynamic model;

[0012] The state error between the simulated state data and the actual state data is determined, and the state error is input into the feedforward neural network to predict the compensation values ​​of each model parameter in the dynamic model;

[0013] Based on the compensation values ​​of each model parameter, the parameters of each model in the dynamic model are adjusted.

[0014] Optionally, the state error is input into a feedforward neural network to predict the compensation values ​​of each model parameter in the dynamic model, specifically including:

[0015] Determine whether the state error is less than a preset threshold;

[0016] If so, determine that the feedforward neural network training is complete, and predict the compensation values ​​of each model parameter in the dynamic model based on the state error and the trained feedforward neural network.

[0017] If not, with the goal of minimizing the input state error, the network parameters in the feedforward neural network are adjusted using the gradient descent method, and the compensation values ​​of each model parameter in the dynamic model are predicted based on the state error and the adjusted feedforward neural network.

[0018] Optionally, when the state error exceeds a preset threshold, the method further includes:

[0019] Based on the compensation values ​​of each model parameter, the parameters of each model in the dynamic model are adjusted;

[0020] Reacquire the historical control data and actual status data of the vehicle during its operation;

[0021] The newly acquired control data is input into the adjusted dynamic model to obtain simulation state data. The state error between the simulation state data and the actual state data is determined until the state error is less than a preset threshold.

[0022] Optionally, with the goal of minimizing the state error of the input, the network parameters in the feedforward neural network are adjusted using gradient descent, specifically including:

[0023] Determine the state error corresponding to each adjustment cycle;

[0024] Based on the changes in state error corresponding to each adjustment cycle, the network parameters in the feedforward neural network are adjusted with the goal of minimizing the input state error.

[0025] Optionally, the state error is input into a feedforward neural network to predict the compensation values ​​of each model parameter in the dynamic model, specifically including:

[0026] The hidden layer node vector is determined based on the state error and the radial basis functions of the hidden layer nodes in the feedforward neural network.

[0027] Based on the hidden layer node vectors and the weights between the hidden layer and the output layer, the compensation values ​​of each model parameter in the dynamic model are predicted.

[0028] Optionally, the feedforward neural network is a radial basis function (RBF) neural network, and the network parameters include the weights between the hidden layer and the output layer, the center vector of the hidden layer nodes, and the basis width parameter.

[0029] Optionally, the model parameters include at least one of the following: transmission ratio of the transmission system, steering ratio of the steering system, braking pressure of the manufacturing system, and throttle opening of the power system.

[0030] This specification provides a dynamic model construction device, including:

[0031] The acquisition module is configured to acquire historical control data and actual status data of the vehicle during operation.

[0032] The simulation module is configured to input the control data into the dynamic model of the vehicle and determine the simulation state data generated by the dynamic model.

[0033] The prediction module is configured to determine the state error between the simulated state data and the actual state data, and input the state error into the feedforward neural network to predict the compensation value of each model parameter in the dynamic model;

[0034] The adjustment module is configured to adjust the parameters of each model in the dynamic model based on the compensation values ​​of each model parameter.

[0035] This specification provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described dynamic model construction method.

[0036] This specification 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 program, it implements the aforementioned dynamic model construction method.

[0037] The above-described at least one technical solution adopted in the embodiments of this specification can achieve the following beneficial effects:

[0038] This specification describes a method for determining the actual state data of a vehicle and the simulated state data generated by a dynamic model based on historical control data from the vehicle's operation. Then, based on the state error between the actual and simulated state data, a feedforward neural network is used to predict the compensation values ​​for each model parameter in the dynamic model. Finally, based on the compensation values ​​of each model parameter, adjustments are made to the dynamic model. By predicting the compensation values ​​of model parameters through a feedforward neural network and subsequently adjusting the dynamic model parameters, automated adjustment of the dynamic model parameters is achieved, shortening the fitting time and improving the fitting efficiency of the dynamic model. Attached Figure Description

[0039] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0040] Figure 1 This document provides a flowchart illustrating a method for constructing a dynamic model, as exemplified in this specification.

[0041] Figure 2 A schematic diagram of the structure of an RBF neural network is provided for the embodiments of this specification;

[0042] Figure 3 A flowchart for adjusting model parameters provided in the embodiments of this specification;

[0043] Figure 4 This is a schematic diagram of the structure of a dynamic model construction device provided in the embodiments of this specification;

[0044] Figure 5 A schematic diagram of an electronic device for implementing the dynamic model construction method provided in the embodiments of this specification. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this specification clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments in this specification without creative effort are within the scope of protection of this application.

[0046] This specification provides a method for constructing a dynamic model. The technical solutions provided by various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0047] Figure 1This is a flowchart illustrating a dynamic model construction method provided in an embodiment of this specification, which may specifically include the following steps:

[0048] S100: Acquire control data and actual status data of the vehicle during its historical operation.

[0049] The dynamic model constructed in this specification is used to fit the motion characteristics of the vehicle, so as to test and verify various driving algorithms in a simulation environment using the fitted dynamic model. The vehicle referred to in this specification refers to vehicles, aircraft, ships, and other equipment, which can be unmanned or manned; this specification makes no limitation on this.

[0050] For ease of description, we will use vehicles as an example for the following explanation.

[0051] To improve the fit between the dynamics model and the vehicle, the parameters of the dynamics model can be adjusted based on the vehicle's driving data under different driving conditions in real road scenarios, so that the adjusted dynamics model better matches the vehicle's motion characteristics.

[0052] Specifically, control data from historical vehicle driving data in road scenarios can be determined, along with actual state data generated based on this control data. Control data can include parameters that drive the vehicle, such as the accelerator pedal opening, brake pedal opening, and steering wheel angle. Actual state data can be divided into actual state quantities and actual perceptual quantities. Actual state quantities characterize the vehicle's operating state and include at least actual speed, actual acceleration, actual angular velocity, actual driving direction, and actual driving trajectory. Actual perceptual quantities characterize changes in the actual state, including the rate of change of actual acceleration and the rate of change of actual angular velocity.

[0053] S102: Input the control data into the dynamic model of the vehicle to determine the simulation state data generated by the dynamic model.

[0054] When constructing the dynamic model in this manual, you can first assign initial values ​​to each model parameter in the dynamic model, and then further adjust the model parameters by comparing the differences between the dynamic model and the vehicle.

[0055] Specifically, the control data can be input into the vehicle's dynamics model. Based on the initial values ​​of each model parameter, the simulation state data generated by the dynamics model is determined. The model parameters include parameters of each functional system in the dynamics model, such as the throttle opening of the powertrain (the opening angle of the engine throttle), the transmission ratio of the transmission system, the steering ratio of the steering system, and the braking pressure of the braking system. The simulation state data can also be divided into simulation state quantities and simulation motion quantities. Simulation state quantities include at least simulation speed, simulation acceleration, simulation angular velocity, and simulation driving direction, while simulation motion quantities include the rate of change of simulation acceleration and the rate of change of simulation angular velocity.

[0056] It should be noted that when comparing the differences between the dynamic model and the vehicle, based on the principle of a single variable, it is also necessary to ensure that the external environmental conditions of the vehicle are the same as the simulation environmental conditions of the dynamic model, such as road conditions and wind speed.

[0057] S104: Determine the state error between the simulated state data and the actual state data, and input the state error into the feedforward neural network to predict the compensation value of each model parameter in the dynamic model.

[0058] S106: Adjust the parameters of each model in the dynamic model according to the compensation values ​​of each model parameter.

[0059] Since the vehicle and the dynamics model are in the same environmental conditions, under the same control conditions, the fitting gap between the dynamics model and the vehicle can be detected based on the difference in their motion states, and then the dynamics model can be adjusted.

[0060] Specifically, under the same control data conditions, the state error between the actual state data of the vehicle and the simulated state data of the dynamic model is determined. This state error is then input into a feedforward neural network to predict the compensation values ​​for various model parameters in the dynamic model. Finally, based on the compensation values ​​of each model parameter, the model parameters in the dynamic model are adjusted. These model parameters include parameters of various functional systems in the dynamic model, such as the throttle opening of the powertrain, the transmission ratio of the transmission system, the steering ratio of the steering system, and the braking pressure of the braking system.

[0061] Furthermore, when determining the state error, for each state dimension, the state error corresponding to that state dimension is determined based on the actual state data and simulation state data under that state dimension. The state dimensions include, but are not limited to, velocity dimension, acceleration dimension, driving direction dimension, driving trajectory dimension, and acceleration rate of change dimension, etc.

[0062] Furthermore, when characterizing the difference in output states between the dynamic model and the vehicle, the error rate of change for each state dimension can be determined, such as the error rate of change for the driving trajectory and the error rate of change for the driving direction. The error rate of change for each state dimension is then synchronously input into the feedforward neural network.

[0063] Furthermore, when predicting compensation values ​​using a feedforward neural network, it is necessary to determine whether the state error between the vehicle and the dynamics model is less than a preset threshold. If the state error is less than the preset threshold, it can be determined that the current feedforward neural network has completed training and has high prediction accuracy. Therefore, based on the input state error and the trained feedforward neural network, the compensation values ​​of each model parameter can be predicted, and the dynamics model can be adjusted based on the compensation values ​​of each model parameter.

[0064] Furthermore, since the state error between the two is less than the preset threshold, it can be considered that the difference between the adaptively adjusted dynamic model and the vehicle is small, and the fit between the two is high. Therefore, the current dynamic model can be used for simulation testing. The preset threshold can be set as needed, and this manual does not impose any restrictions on it.

[0065] When the state error between the two exceeds a preset threshold, it is considered that the current feedforward neural network still needs training, and the dynamic model also needs further adjustment. Therefore, with the goal of minimizing the input state error, the network parameters in the feedforward neural network can be adjusted using gradient descent, and the compensation values ​​of each model parameter can be predicted using the adjusted feedforward neural network and the state error.

[0066] Subsequently, the model parameters of the dynamic model are adjusted based on the compensation values ​​of each model parameter predicted by the adjusted feedforward neural network.

[0067] Then, the control data and actual state data from the vehicle's historical operation are reacquired. The reacquired control data is then input into the adjusted dynamic model to obtain simulation state data. The state error between the simulation state data and the actual state data is determined, and the training is considered complete when the state error is less than a preset threshold.

[0068] In order to adapt to the vehicle's motion characteristics under various driving conditions, the newly acquired control data can be selected from the historically acquired control data under different driving conditions or at different driving times.

[0069] Furthermore, during each adjustment cycle, when adjusting the network parameters in the feedforward neural network, the state errors corresponding to each scheduling cycle before that cycle can be determined. Based on the changes in the previous state errors, the network parameters in the feedforward neural network are adjusted with the goal of minimizing the state error of the current input. For example, if reducing the weights between the hidden layer and the output layer reduces the state error between them, the weights between the hidden layer and the output layer can be further reduced.

[0070] The network parameters include the center vector of the hidden layer nodes, the base width parameter, and the weights between the hidden layer and the output layer.

[0071] In the embodiments of this specification, the feedforward neural network can be a common BP (Back Propagation) neural network, a radial basis function (RBF) neural network, etc. Among them, the RBF neural network has strong generalization ability, faster adaptability and self-learning ability, and higher prediction accuracy. Therefore, the RBF neural network model can be preferred for prediction to improve the accuracy of the fitting between the dynamic model and the vehicle.

[0072] Figure 2 This is a schematic diagram of the RBF neural network provided in the embodiments of this specification. The RBF neural network includes a three-layer network structure: an input layer, hidden layers, and an output layer. The state errors x1, x2…xn of each state dimension correspond to the input data of each node in the input layer. Based on the input state errors, the input data is mapped from a low-dimensional space to a high-dimensional space through the radial basis functions of each hidden layer node, resulting in the hidden layer node vectors. Finally, based on the hidden layer node vectors and the weights between the hidden layer and the output layer, the compensation values ​​for each model parameter in the final output are determined.

[0073] In the RBF neural network, the number of input layer nodes can be determined based on the state dimension of the input state error, the number of output layer nodes can be determined based on the model parameters to be adjusted, and the number of hidden layer nodes can be determined based on at least one of the number of input layer nodes or the number of output layer nodes.

[0074] For example, when determining the number of hidden layer nodes, any one of the following three methods can be used.

[0075]

[0076] m = log₂n

[0077]

[0078] In the above formula, m represents the number of hidden layer nodes, n represents the number of input layer nodes, s represents the number of output layer nodes, and α is a constant between 1 and 10.

[0079] In this RBF neural network, for any node in the hidden layer, the hidden node vector output by that hidden layer node is:

[0080]

[0081] Where j represents the j-th node in the hidden layer, and X = [x1, x2, ..., xn] T The input is an n-dimensional state error, where n represents the number of state dimensions, and C... j =[C j1 C j2 …C jn ] T Let σ represent the center vector of the j-th hidden node. j This represents the base width parameter of the j-th hidden layer node.

[0082] Then, based on the hidden node vectors output by each hidden node and the weights between each hidden node and the output node, the compensation values ​​for each model parameter are determined:

[0083]

[0084] Among them, O k w represents the compensation value of the model parameters output by the k-th node in the output layer. jk This represents the weight between the j-th hidden layer node and the k-th output layer node.

[0085] It should be noted that when adaptively adjusting parameters using an RBF neural network, initial values ​​can be assigned to the center vector of the hidden layer nodes, the base width parameter, and the weights between the hidden and output layers. In subsequent steps, the network parameters can be adjusted based on changes in the state error.

[0086] Figure 3 The flowchart for model parameter adjustment provided in the embodiments of this specification allows inputting historical vehicle control data in road scenarios into the dynamics model to obtain simulated state data. This simulated state data is then compared with the actual state data generated by the vehicle to determine the state error between the two. Subsequently, a radial basis function neural network is used to predict the compensation values ​​of each model parameter in the dynamics model, and compensation is applied to each model parameter. If the state error exceeds a preset threshold, new control data is acquired, and the above steps are repeated to determine the compensation values ​​of each model parameter, continuously adjusting the model parameters in the dynamics model until the state error is less than the preset threshold. This model parameter adjustment process achieves closed-loop automated model parameter adjustment, overcoming the shortcomings of manual parameter adjustment.

[0087] based on Figure 1The illustrated method for constructing a dynamic model determines the actual state data of the vehicle and the simulated state data generated by the dynamic model based on historical control data from the vehicle's operation. Then, based on the state error between the actual and simulated state data, a feedforward neural network predicts the compensation values ​​for each model parameter in the dynamic model. Finally, based on the compensation values, the model parameters in the dynamic model are adjusted. By predicting the compensation values ​​of the model parameters through a feedforward neural network and then adjusting the parameters of the dynamic model, automated adjustment of the dynamic model parameters is achieved, shortening the fitting time of the dynamic model and improving the fitting efficiency and iterative update efficiency.

[0088] based on Figure 1 The present invention provides a method for constructing a dynamic model, and the embodiments of this specification also provide a structural schematic diagram of a dynamic model construction device, as shown in the figure. Figure 4 As shown.

[0089] Figure 4 A schematic diagram of a dynamic model building device provided in the embodiments of this specification includes:

[0090] The acquisition module 200 is configured to acquire historical control data and actual status data of the vehicle during operation.

[0091] The simulation module 202 is configured to input the control data into the dynamic model of the vehicle and determine the simulation state data generated by the dynamic model.

[0092] The prediction module 204 is configured to determine the state error between the simulation state data and the actual state data, and input the state error into the feedforward neural network to predict the compensation value of each model parameter in the dynamic model;

[0093] The adjustment module 206 is configured to adjust each model parameter in the dynamic model according to the compensation value of each model parameter.

[0094] Optionally, the prediction module 204 is specifically used to determine whether the state error is less than a preset threshold. If so, it determines that the feedforward neural network training is complete, and predicts the compensation values ​​of each model parameter in the dynamic model based on the state error and the trained feedforward neural network. If not, it adjusts the network parameters in the feedforward neural network using gradient descent with the goal of minimizing the input state error, and predicts the compensation values ​​of each model parameter in the dynamic model based on the state error and the adjusted feedforward neural network.

[0095] Optionally, when the state error is greater than a preset threshold, the prediction module 204 is further configured to adjust each model parameter in the dynamic model according to the compensation value of each model parameter, re-acquire the control data and actual state data of the vehicle during its historical operation, input the re-acquired control data into the adjusted dynamic model to obtain simulation state data, determine the state error between the simulation state data and the actual state data, until the state error is less than the preset threshold.

[0096] Optionally, the prediction module 204 is specifically used to determine the state error corresponding to each adjustment period in history, and adjust the network parameters in the feedforward neural network based on the changes in the state error corresponding to each adjustment period, with the goal of minimizing the input state error.

[0097] Optionally, the prediction module 204 is specifically used to determine the hidden layer node vector based on the state error and the radial basis function of the hidden layer node in the feedforward neural network, and to predict the compensation value of each model parameter in the dynamic model based on the hidden layer node vector and the weight between the hidden layer and the output layer.

[0098] Optionally, the feedforward neural network is a radial basis function (RBF) neural network, and the network parameters include the weights between the hidden layer and the output layer, the center vector of the hidden layer nodes, and the basis width parameter.

[0099] Optionally, the model parameters include at least one of the following: transmission ratio of the transmission system, steering ratio of the steering system, braking pressure of the manufacturing system, and throttle opening of the power system.

[0100] This specification also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described embodiments. Figure 1 The provided method for constructing dynamic models.

[0101] according to Figure 1 The present invention provides a method for constructing a dynamic model, and the embodiments of this specification also propose... Figure 5 The diagram shows a schematic structural representation of the electronic device. Figure 5 At the hardware level, this electronic device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then executes it to achieve the above. Figure 1 The method for constructing the dynamic model is shown.

[0102] Of course, in addition to software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of hardware and software. In other words, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0103] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using hardware physical modules. For example, a Programmable Logic Device (PLD) (e.g., a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Moreover, nowadays, instead of manually generating integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0104] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0105] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0106] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware components.

[0107] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0108] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0109] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0110] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0111] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0112] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0113] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0114] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0115] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0116] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0117] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0118] The above description is merely an embodiment of this specification and is not intended to limit this specification. Various modifications and variations can be made to this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of the claims of this specification.

Claims

1. A method for constructing a dynamic model, characterized in that, include: Acquire historical control data and actual status data of the vehicle during operation; The control data is input into the dynamic model of the vehicle to determine the simulation state data generated by the dynamic model; The state error between the simulated state data and the actual state data is determined, and the state error is input into the feedforward neural network to predict the compensation values ​​of each model parameter in the dynamic model; Based on the compensation values ​​of each model parameter, the parameters of each model in the dynamic model are adjusted; The state error is input into a feedforward neural network to predict the compensation values ​​of each model parameter in the dynamic model, specifically including: Determine whether the state error is less than a preset threshold; If so, determine that the feedforward neural network training is complete, and predict the compensation values ​​of each model parameter in the dynamic model based on the state error and the trained feedforward neural network. If not, with the goal of minimizing the input state error, the network parameters in the feedforward neural network are adjusted using the gradient descent method, and the compensation values ​​of each model parameter in the dynamic model are predicted based on the state error and the adjusted feedforward neural network.

2. The method as described in claim 1, characterized in that, When the state error is greater than a preset threshold, the method further includes: Based on the compensation values ​​of each model parameter, the parameters of each model in the dynamic model are adjusted; Reacquire the historical control data and actual status data of the vehicle during its operation; The newly acquired control data is input into the adjusted dynamic model to obtain simulation state data. The state error between the simulation state data and the actual state data is determined until the state error is less than a preset threshold.

3. The method as described in claim 2, characterized in that, With the goal of minimizing the input state error, the network parameters in the feedforward neural network are adjusted using gradient descent, specifically including: Determine the state error corresponding to each adjustment cycle in history; Based on the changes in state error corresponding to each adjustment cycle, the network parameters in the feedforward neural network are adjusted with the goal of minimizing the input state error.

4. The method as described in claim 1, characterized in that, The state error is input into a feedforward neural network to predict the compensation values ​​of each model parameter in the dynamic model, specifically including: The hidden layer node vector is determined based on the state error and the radial basis functions of the hidden layer nodes in the feedforward neural network. Based on the hidden layer node vectors and the weights between the hidden layer and the output layer, the compensation values ​​of each model parameter in the dynamic model are predicted.

5. The method as described in any one of claims 1 or 3, characterized in that, The feedforward neural network is a radial basis function (RBF) neural network, and the network parameters include the weights between the hidden layer and the output layer, the center vector of the hidden layer nodes, and the basis width parameter.

6. The method as described in claim 1, characterized in that, The model parameters include at least one of the following: transmission ratio of the transmission system, steering ratio of the steering system, braking pressure of the manufacturing system, and throttle opening of the power system.

7. A dynamic model construction device, characterized in that, include: The acquisition module is configured to acquire historical control data and actual status data of the vehicle during operation. The simulation module is configured to input the control data into the dynamic model of the vehicle and determine the simulation state data generated by the dynamic model. The prediction module is configured to determine the state error between the simulated state data and the actual state data, and input the state error into the feedforward neural network to predict the compensation value of each model parameter in the dynamic model; The adjustment module is configured to adjust the parameters of each model in the dynamic model based on the compensation values ​​of each model parameter. The state error is input into a feedforward neural network to predict the compensation values ​​of each model parameter in the dynamic model, specifically including: Determine whether the state error is less than a preset threshold; If so, determine that the feedforward neural network training is complete, and predict the compensation values ​​of each model parameter in the dynamic model based on the state error and the trained feedforward neural network. If not, with the goal of minimizing the input state error, the network parameters in the feedforward neural network are adjusted using the gradient descent method, and the compensation values ​​of each model parameter in the dynamic model are predicted based on the state error and the adjusted feedforward neural network.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method described in any one of claims 1 to 6.

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 program, it implements the method described in any one of claims 1 to 6.

Images