Method for generating a vehicle longitudinal control model for an artificial intelligence-based system for longitudinal control of autonomous vehicles, computing device, and computer program

Abstract

This invention provides a method for generating a vehicle longitudinal control model for an artificial intelligence-based system for longitudinal control of autonomous vehicles, a computing device, and a computer program. [Solution] A method for generating a vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of an autonomous vehicle, which is performed by a computing device, includes the steps of: acquiring driving logging data for a vehicle; generating training data using the acquired driving logging data; and generating a vehicle longitudinal control model that derives result data for vehicle longitudinal control using the generated training data.

Description

Technical Field

[0001] Various embodiments of the present disclosure relate to a method for generating a vehicle longitudinal control model of an artificial intelligence infrastructure for longitudinal control of an autonomous vehicle, a computing device, and a computer program.

Background Art

[0002] For the convenience of users driving vehicles, various sensors and electronic devices (e.g., there is a trend of being equipped with a vehicle driver assistance system (ADAS: Advanced Driver Assistance System)), especially, without driver intervention, recognizing the surrounding environment, and the development of technologies for an autonomous driving system of a vehicle that automatically travels to a given destination according to the recognized surrounding environment has been actively carried out.

[0003] An autonomous vehicle refers to a vehicle equipped with an autonomous driving system function that recognizes the surrounding environment without driver intervention and automatically travels to a given destination according to the recognized surrounding environment, and the autonomous driving system function means performing lateral position, perception, prediction, planning, and control for autonomous driving.

[0004] The autonomous driving system processes point cloud data acquired through sensors (e.g., lidar sensors) through a recognition process to detect objects located around the autonomous vehicle, and through a planning process, the recognition results (e.g., information such as the position, posture, and speed of the object) derived by performing the recognition process are transmitted to establish a driving plan such as the route and speed of the autonomous vehicle.

[0005] The above-mentioned background art is what the inventor possessed or acquired during the process of deriving the content of the present disclosure, and it is not necessarily prior art publicly disclosed to the general public before this application.

Summary of the Invention

Problems to be Solved by the Invention

[0006] To design a system for longitudinal vehicle control (e.g., speed control) within an autonomous driving system, it is necessary to mathematically model the vehicle's dynamics.

[0007] However, modeling the dynamics of a vehicle is difficult because it must accurately reflect the vehicle's complex physical characteristics and the influence of the external environment. For example, vehicle dynamics are nonlinearly influenced by various factors such as velocity, acceleration, and resistance, and modeling them requires considering the engine, transmission, tire friction, air resistance, and other factors, resulting in a highly complex model.

[0008] Traditionally, the dynamics models of such vehicles have been solved through differential equation modeling. However, this method has the problem that determining the parameters of the differential equations requires a great deal of time, expense, and effort. For example, determining the parameters of the differential equations requires acquiring actual vehicle driving data and mapping the correlations of this data one by one, a process that requires considerable time resources and effort.

[0009] Therefore, the problem that this disclosure aims to solve is to provide an artificial intelligence-based vehicle longitudinal control model generation method, computing device, and computer program for the longitudinal control of autonomous vehicles that can design the vehicle longitudinal control model more efficiently by constructing an artificial intelligence-based vehicle longitudinal control model that can derive information for longitudinal control of the vehicle based on the vehicle's target acceleration and vehicle state information, and by performing longitudinal control of the vehicle based on the resulting data derived from this model.

[0010] The problems that this disclosure seeks to solve are not limited to those mentioned above, and any other problems not mentioned can be clearly understood by an ordinary engineer from the following description. [Means for solving the problem]

[0011] A method for generating a vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of an autonomous vehicle, according to one embodiment of the present disclosure for solving the aforementioned problems, is a method for generating a vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of an autonomous vehicle, performed by a computing device, and may include the steps of: acquiring driving logging data for a vehicle; generating training data using the acquired driving logging data; and generating a vehicle longitudinal control model that derives result data for longitudinal control of the vehicle using the generated training data.

[0012] In various embodiments, the step of generating the learning data may include the step of obtaining a target acceleration from the acquired driving logging data, the step of obtaining vehicle state information from the acquired driving logging data, the step of obtaining acceleration pedal position information from the acquired driving logging data, and the step of generating learning data that includes the acquired target acceleration and the acquired vehicle state information as input data, and the acquired acceleration pedal position information as ground truth data.

[0013] In various embodiments, the step of obtaining the target acceleration may include a step of obtaining the acceleration measured from the vehicle at a second time point, which is a predetermined time after the first time point, as the target acceleration at the first time point.

[0014] In various embodiments, the step of obtaining the target acceleration may include a step of obtaining the average value of accelerations measured from the vehicle over a predetermined period of time after a specific point in time as the target acceleration at that specific point in time.

[0015] In various embodiments, the step of acquiring the vehicle status information may include the step of acquiring information regarding the vehicle's longitudinal speed, engine RPM, and gear status from the acquired driving logging data as vehicle status information.

[0016] In various embodiments, the step of acquiring the vehicle state information may include a step of acquiring at least one of the following information from the acquired driving logging data: the vehicle's tilt information and the amount of change in the vehicle's tilt.

[0017] In various embodiments, the step of generating the learning data includes the step of acquiring driving environment information while the vehicle is driving based on pre-generated precision map data, and the step of generating learning data using the acquired driving environment information and the acquired driving logging data, wherein the acquired driving environment information may include at least one of the gradient and the change in gradient of the location where the vehicle is driving.

[0018] In various embodiments, the step of generating training data using the acquired driving environment information and the acquired driving logging data may include the step of acquiring at least one of the vehicle's tilt information and tilt change amount information as vehicle state information, based on at least one of the inclination and change in inclination of the location where the vehicle is driving, and the step of generating training data using the acquired at least one piece of information and the acquired driving logging data.

[0019] In various embodiments, the generated vehicle longitudinal control model may be an STM (Long Short-Term Memory) model.

[0020] In various embodiments, the step of generating the learning data includes the steps of obtaining a target acceleration from the acquired driving logging data, obtaining vehicle state information from the acquired driving logging data, extracting an estimated acceleration corresponding to the acquired target acceleration and the acquired vehicle state information, and generating learning data including the acquired target acceleration and the acquired vehicle state information as input data and including the extracted estimated acceleration as correct answer data. The step of generating the vehicle longitudinal control model may include the step of generating a vehicle longitudinal control model that derives a combination of optimal parameter functions that constitute a vehicle longitudinal model derived from a vehicle dynamics model by learning using the generated learning data.

[0021] In various embodiments, the step of generating a vehicle longitudinal control model that derives a combination of optimal parameter functions that constitute the vehicle longitudinal model may include the step of generating a plurality of neural network models as unit models for each of the plurality of parameter functions when the vehicle longitudinal model includes a plurality of parameter functions, and the step of generating one vehicle longitudinal control model including the generated plurality of neural network models.

[0022] In various embodiments, the vehicle dynamics model may be expressed as shown in the following Mathematical Formula 1.

[0023] <Mathematical Formula 1>

Number

[0024] Here, the TIFF2026106386000003.tif66 is the engine output, the TIFF2026106386000004.tif610 is the rolling resistance, the TIFF2026106386000005.tif1013 is the air resistance, the TIFF2026106386000006.tif99 is the boarding resistance and the TIFF2026106386000007.tif99 may be the inertial resistance.

[0025] In various embodiments, the vehicle longitudinal model may be expressed as in the following mathematical formula 2.

[0026] <Mathematical formula 2> [Number]

[0027] Here, the TIFF2026106386000009.tif54 is the vehicle longitudinal model, the TIFF2026106386000010.tif59 is a parameter function related to the gear scaling coefficient according to the gear state, the TIFF2026106386000011.tif65 is a static parameter related to the gear, the TIFF2026106386000012.tif54 is the gear state, the TIFF2026106386000013.tif711 is a gear parameter function according to the speed, the TIFF2026106386000014.tif43 is the speed, the TIFF2026106386000015.tif515 is the accelerator pedal position value and the TIFF2026106386000016.tif725 may be a DC gain function that varies according to the accelerator pedal position value and the gear state.

[0028] In various embodiments, the derived combination of optimal parameter functions may include a parameter function related to the gear scaling coefficient according to the gear state, a parameter function related to the gear, and a parameter function related to the DC gain that varies according to the accelerator pedal position value and the gear state.

[0029] In various embodiments, the step of generating the training data may include a step of preprocessing the acquired driving logging data and a step of generating training data using the preprocessed driving logging data.

[0030] A computing device for performing a method for generating a vehicle longitudinal control model for an artificial intelligence-based vehicle longitudinal control, according to another embodiment of the present disclosure for solving the aforementioned problems, includes a processor, a network interface, memory, and a computer program loaded into the memory and executed by the processor, wherein the computer program may include instructions for acquiring driving logging data for a vehicle, instructions for generating training data using the acquired driving logging data, and instructions for generating a vehicle longitudinal control model that derives result data for vehicle longitudinal control using the generated training data.

[0031] A computer program according to yet another embodiment of the present disclosure for solving the aforementioned problems may be stored on a recording medium readable by a computing device in conjunction with the computing device to execute a method for generating a vehicle longitudinal control model for the longitudinal control of an autonomous vehicle, which includes the steps of: acquiring driving logging data for a vehicle; generating training data using the acquired driving logging data; and generating a vehicle longitudinal control model that derives result data for vehicle longitudinal control using the generated training data.

[0032] Further specific details of this disclosure are included in the detailed description and drawings. [Effects of the Invention]

[0033] According to the various embodiments of this disclosure, an artificial intelligence-based vehicle longitudinal control model can be constructed that can derive information for vehicle longitudinal control based on the vehicle's target acceleration and vehicle state information, and vehicle longitudinal control can be performed based on the resulting data derived from this model, which has the advantage of allowing for more efficient design of the vehicle longitudinal control model.

[0034] The effects of this disclosure are not limited to those mentioned above, and any other effects not mentioned can be clearly understood by an ordinary person in the art from the following description. [Brief explanation of the drawing]

[0035] The following drawings accompanying this specification illustrate preferred embodiments of the Disclosure and, together with the detailed description of the invention, serve to further illustrate the technical concept of the Disclosure; therefore, this Disclosure should not be construed as being limited solely to what is depicted in such drawings. [Figure 1] This is a diagram illustrating an autonomous driving system according to one embodiment of the present disclosure. [Figure 2] This is a diagram illustrating the hardware configuration of a computing device according to another embodiment of the present disclosure. [Figure 3] This is a flowchart of a first vehicle longitudinal control model generation method for an artificial intelligence platform for longitudinal control of autonomous vehicles, in various embodiments. [Figure 4] This diagram illustrates pre-processed driving logging data applicable to various embodiments. [Figure 5] This is a diagram illustrating a first vehicle longitudinal control model applicable to various embodiments. [Figure 6] This is a schematic diagram showing a network function related to one embodiment of the present disclosure. [Figure 7] This diagram illustrates an LSTM (Long Short-Term Memory) model applicable to various implementations. [Figure 8] This diagram illustrates the learning results of vehicle longitudinal control models for various embodiments. [Figure 9] This diagram illustrates the learning results of vehicle longitudinal control models for various embodiments. [Figure 10] This diagram illustrates the experimental results of a vehicle longitudinal control model learned through various implementations. [Figure 11] This is a flowchart of a method for generating a second vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of autonomous vehicles, in various embodiments. [Figure 12] This is a diagram illustrating a second vehicle longitudinal control model applicable to various embodiments. [Figure 13] This drawing illustrates a vehicle longitudinal model applicable to various embodiments. [Modes for carrying out the invention]

[0036] The advantages and features of this disclosure, and how they are achieved, will become clearer with reference to the embodiments described below in detail with the accompanying drawings. However, this disclosure is not limited to the embodiments disclosed below and may be embodied in a variety of different forms, although these embodiments are provided to complete the disclosure and to fully inform a person ordinary in the art to which this disclosure belongs of the scope of this disclosure, which is defined only by the scope of the claims.

[0037] The terms used herein are for illustrative purposes only and are not intended to limit the disclosure. In this specification, singular terms include plural terms unless otherwise specified. The terms “comprises” and / or “comprising” as used in this specification do not preclude the presence or addition of one or more other components in addition to those mentioned.

[0038] Throughout this specification, the same drawing reference numerals refer to the same component, and "and / or" includes each of the components mentioned and all combinations of one or more of them. Although terms such as "first," "second," etc., are used to describe a variety of components, these components are not limited by these terms. These terms are simply used to distinguish one component from another. Therefore, the first component mentioned below may, of course, be the second component within the technical concept of this disclosure.

[0039] As used herein, the terms “part” or “module” refer to software, hardware components such as FPGAs or ASICs, and “part” or “module” perform a role. However, “part” or “module” is not limited to software or hardware. A “part” or “module” may be configured to reside on an addressable storage medium, or to regenerate one or more processors. Thus, as an example, a “part” or “module” includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, processors, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. Components and the functions provided within “parts” or “modules” may be combined with a smaller number of components and “parts” or “modules,” or further separated into “parts” or “modules” with additional components.

[0040] Spatially relative terms such as "below," "beneath," "lower," "above," and "upper" can be used to easily describe the correlation between one component and another, as illustrated in the drawing. Spatially relative terms should be understood as including not only the directions illustrated in the drawing, but also the different directions of components during use or operation. For example, if a component illustrated in the drawing is flipped over, a component described as "below" or "beneath" of another component may be placed "above" of that other component. Therefore, the illustrative term "below" can include both downward and upward directions. Components can also be oriented in other directions, and accordingly, spatially relative terms can be interpreted in terms of orientation.

[0041] In this specification, expressions such as "first," "second," "initial," and "second" are used to distinguish one object from others when referring to multiple similar objects, unless the context indicates otherwise, and do not limit the order or importance of the objects in question.

[0042] Expressions used herein such as “A, B, and C,” “A, B, or C,” “A, B, and / or C,” or “at least one of A, B, and C,” “at least one of A, B, or C,” “at least one of A, B, and / or C,” “at least one selected from A, B, and C,” “at least one selected from A, B, or C,” and “at least one selected from A, B, and / or C” may mean each listed item or all possible combinations of the listed items. For example, “at least one selected from A and B” may refer to (1) A, (2) at least one of A, (3) B, (4) at least one of B, (5) at least one of A and at least one of B, (6) at least one of A and B, (7) at least one of B and A, and (8) A and B.

[0043] As used herein, the expression "based on" is used to describe one or more factors that influence an act or action of decision, judgment, or action described in the phrase or sentence containing the expression, and this expression does not exclude any additional factors that influence the act or action of decision, judgment, or action.

[0044] As used herein, the expressions "connected" or "linked" of one component (e.g., component 1) to another component (e.g., component 2) may mean not only that the component is directly connected or linked to the other component, but also that it is connected or linked through a new other component (e.g., component 3).

[0045] The expression "configured to" as used herein can mean, depending on the context, "set to do," "capable of doing," "modified to do," "made to do," or "capable of doing." The expression is not limited to the meaning of "specifically designed in hardware," and for example, a processor configured to perform a specific operation may mean a generic-purpose processor that can perform that specific operation by running software.

[0046] Unless otherwise defined, all terms used herein (including technical and scientific terms) should be used in a sense that can be commonly understood by a person of ordinary skill in the art to which this disclosure pertains. Furthermore, terms defined in commonly used dictionaries should not be interpreted ideally or excessively unless explicitly defined otherwise.

[0047] In this specification, "computer" means all types of hardware devices including at least one processor, and may be understood, depending on the embodiment, to also include software configurations running on such hardware devices. For example, "computer" may be understood to include, but is not limited to, smartphones, tablet PCs, desktops, laptops, and all user clients and applications that run on each device.

[0048] The embodiments of this disclosure will be described in detail below with reference to the attached drawings.

[0049] Although each step described herein is described as being performed by a computer, the main body of each step is not limited thereto, and in some embodiments, at least part of each step may be performed by different devices.

[0050] Figure 1 is a diagram illustrating an autonomous driving system according to one embodiment of the present disclosure.

[0051] Referring to Figure 1, an autonomous driving system according to one embodiment of the present invention may include a computing device 100, a user terminal 200, an external server 300, and a network 400.

[0052] Here, the autonomous driving system shown in Figure 1 is based on one embodiment, and its components are not limited to the embodiment shown in Figure 1, but may be added, modified, or deleted as needed.

[0053] In one embodiment, the computing device 100 can perform autonomous driving control for the vehicle 10. To this end, the computing device 100 can perform lateral movement, cognitive movement, planning movement, and control movement.

[0054] In the autonomous driving system shown in Figure 1, the computing device 100 may be located outside the vehicle 10. The computing device 100 determines control commands related to autonomous driving outside the vehicle 10 and transmits these commands to the vehicle 10, thereby enabling the vehicle 10 to perform autonomous driving operations. However, the system is not limited to this; the computing device 100 may also be one of the components located inside the vehicle 10. The computing device 100 determines control commands related to autonomous driving inside the vehicle 10 and directly controls the components of the vehicle 10 using these commands, thereby enabling autonomous driving control of the vehicle 10. For example, the computing device 100 may be a control module that controls the operation of components included in the vehicle 10 from within the vehicle 10.

[0055] More specifically, the lateral movement performed by the computing device 100 may mean the movement of measuring the position and attitude of the vehicle 10. For example, the computing device 100 can collect sensor data (e.g., point cloud data, image data, etc.) by scanning the surrounding environment of the vehicle 10 using sensors installed in the vehicle 10 (e.g., LiDAR, RADAR, Camera, GNSS / INS, IMU, etc.), and can use the collected sensor data to derive lateral information, including lateral values ​​corresponding to the position and attitude of the vehicle 10.

[0056] Next, the cognitive operations performed by the computing device 100 may involve detecting and tracking objects located around the vehicle 10. For example, the computing device 100 can recognize objects present around the vehicle 10 by analyzing sensor data (e.g., LiDAR data and image data) collected by scanning the area around the vehicle 10, and can derive cognitive information containing information about the recognized objects.

[0057] Next, the planning operation performed by the computing device 100 may mean the operation of generating a driving plan, including a path and speed profile for controlling the vehicle 10, based on the lateral information derived through the lateral module and the cognitive information derived through the cognitive module.

[0058] Finally, the control operations performed by the computing device 100 can determine and generate control commands for lateral control (direction control) and longitudinal control (speed control) of the vehicle 10 based on information about the travel plan derived from the planning operations, and the operation of the vehicle 10 can be controlled by the determined and generated control commands.

[0059] In various embodiments, the computing device 100 can derive result data for longitudinal control of the vehicle based on a vehicle longitudinal control model, and perform longitudinal control of the vehicle 10 based on this. To this end, the computing device 100 can generate training data based on driving logging data for the vehicle 10, and generate a vehicle longitudinal control model that derives result data for longitudinal control of the vehicle by training it based on this data.

[0060] In various embodiments, the computing device 100 can be connected to a user terminal 200 via a network 400 and can provide the user terminal 200 with various information related to autonomous driving.

[0061] Here, user terminal 200 can mean any form of entity(s) in a system having a mechanism for communication with computing device 100. For example, such user terminal 200 can include PCs (personal computers), notebooks, mobile terminals, smartphones, tablet PCs, and wearable devices, and can include all types of terminals that can connect to wired / wireless networks. User terminal 200 may also include any computing device embodied by at least one of an agent, an API (Application Programming Interface), and a plug-in. User terminal 200 may also include an application source and / or client application.

[0062] Furthermore, the network 400 here may refer to a connected structure that enables information exchange between nodes such as multiple terminals and servers. For example, the network 400 may include local area networks (LANs), wide area networks (WANs), the internet (WWW), wired / wireless data networks, telephone networks, wired / wireless television networks, controller area networks (CANs), and Ethernet.

[0063] Wireless data communication networks include, but are not limited to, 3G, 4G, 5G, 3GPP (3rd Generation Partnership Project), 5GPP (5th Generation Partnership Project), LTE (Long Term Evolution), WiMAX (World Interoperability for Microwave Access), Wi-Fi, Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), RF (Radio Frequency), Bluetooth networks, NFC (Near-Field Communication) networks, satellite broadcasting networks, analog broadcasting networks, and DMB (Digital Multimedia Broadcasting) networks.

[0064] In one embodiment, the external server 300 may be connected to the computing device 100 via a network 400, and can store and manage various information and data necessary for the computing device 100 to perform the vehicle longitudinal control model generation method of the artificial intelligence base for longitudinal control of autonomous vehicles, or can collect, store and manage various information and data derived by the computing device 100 performing the vehicle longitudinal control model generation method of the artificial intelligence base for longitudinal control of autonomous vehicles. For example, the external server 300 may be, but is not limited to, a storage server separately provided outside the computing device 100. Hereinafter, with reference to Figure 2, the hardware configuration of the computing device 100 that performs the vehicle longitudinal control model generation method of the artificial intelligence base for longitudinal control of autonomous vehicles will be described.

[0065] Figure 2 is a diagram illustrating the hardware configuration of a computing device according to another embodiment of the present disclosure.

[0066] Referring to Figure 2, a computing device 100 according to another embodiment of the present disclosure may include one or more processors 110, memory 120 for loading computer programs 151 executed by the processors 110, a bus 130, a communication interface 140, and storage 150 for storing the computer programs 151. Here, only components relevant to the embodiments of the present disclosure are illustrated in Figure 2. Therefore, a person of the ordinary skill in the art to which the present disclosure belongs will see that other general-purpose components may be included in addition to those illustrated in Figure 2.

[0067] The processor 110 controls the overall operation of each component of the computing device 100. The processor 110 may include a CPU (Central Processing Unit), an MPU (Micro Processor Unit), an MCU (Micro Controller Unit), a GPU (Graphics Processing Unit), or any form of processor widely known in the art of this disclosure.

[0068] Furthermore, the processor 110 can perform calculations for at least one application or program to carry out the method according to the embodiments of this disclosure, and the computing device 100 may comprise one or more processors.

[0069] In various embodiments, the processor 110 may further include Random Access Memory (RAM) (not shown) and Read-Only Memory (ROM) (not shown) for temporarily and / or permanently storing signals (or data) processed within the processor 110. The processor 110 may also be implemented as a System on Chip (SoC) including at least one of the graphics processing unit, RAM, and ROM.

[0070] Memory 120 stores various data, instructions, and / or information. Memory 120 can load a computer program 151 from storage 150 to perform a method / operation according to various embodiments of the present disclosure. Once the computer program 151 is loaded into memory 120, the processor 110 can perform the method / operation by executing one or more instructions that constitute the computer program 151. Memory 120 may be embodied in volatile memory such as RAM, but the technical scope of the present disclosure is not limited thereto.

[0071] Bus 130 provides communication functions between components of the computing device 100. Bus 130 can be implemented in various forms, such as an address bus, a data bus, and a control bus.

[0072] The communication interface 140 supports wired / wireless internet communication of the computing device 100. The communication interface 140 may also support a variety of communication methods other than internet communication. For this purpose, the communication interface 140 may be configured to include communication modules widely known in the art of this disclosure. In some embodiments, the communication interface 140 may be omitted.

[0073] The storage 150 can temporarily store the computer program 151. When the process of generating a vehicle longitudinal control model for the artificial intelligence base for the longitudinal control of an autonomous vehicle is performed through the computing device 100, the storage 150 can store various information necessary to provide the process of generating a vehicle longitudinal control model for the artificial intelligence base for the longitudinal control of an autonomous vehicle.

[0074] The storage 150 may consist of non-volatile memory such as ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), flash memory, a hard disk, a removable disk, or any form of computer-readable recording medium widely known in the art to which this disclosure belongs.

[0075] When the computer program 151 is loaded into memory 120, it may include one or more instructions that cause the processor 110 to perform the methods / operations according to various embodiments of the present disclosure. That is, the processor 110 can perform the methods / operations according to various embodiments of the present disclosure by executing the one or more instructions.

[0076] In one embodiment, the computer program 151 may include one or more instructions to perform a method for generating a vehicle longitudinal control model on an artificial intelligence platform for longitudinal control of an autonomous vehicle, which includes the steps of acquiring driving logging data for a vehicle, generating training data using the acquired driving logging data, and generating a vehicle longitudinal control model that derives result data for vehicle longitudinal control using the generated training data.

[0077] Steps of the methods or algorithms described in relation to the embodiments of this disclosure may be embodied directly in hardware, in software modules executed by hardware, or in combination thereof. The software modules may reside on RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), flash memory, hard disk, removable disk, CD-ROM, or any form of computer-readable recording medium widely known in the art to which this disclosure belongs.

[0078] The components of this disclosure may be embodied in a program (or application) and stored on a medium for execution in conjunction with a computer, which is hardware. The components of this disclosure may be executed by software programming or software elements, and similarly, embodiments include a variety of algorithms embodied in combinations of data structures, processes, routines or other programming configurations, which may be embodied in programming or scripting languages ​​such as C, C++, Java, and assembler. Functional aspects may be embodied in algorithms executed on one or more processors. The following describes a method for generating a vehicle longitudinal control model for an artificial intelligence base for longitudinal control of an autonomous vehicle, performed by a computing device 100, with reference to Figures 3 to 13.

[0079] Figure 3 is a flowchart of the first vehicle longitudinal control model generation method for an artificial intelligence platform for longitudinal control of autonomous vehicles in various embodiments.

[0080] Referring to Figure 3, at step S110, the computing device 100 can acquire driving logging data for the vehicle 10.

[0081] Here, the driving logging data for vehicle 10 is data collected through various sensors installed in vehicle 10 while vehicle 10 is in motion, and may be data recording vehicle 10's operation and environmental information. For example, the driving logging data may include vehicle 10's operation and test environment information collected during the process of vehicle 10 performing a test run, or it may include vehicle 10's operation and actual environmental information collected during the process of vehicle 10 actually performing a run.

[0082] As an example, the driving logging data may include data collected through the ECU (Electronic Control Unit) (e.g., control signal data such as the engine, gear status, and acceleration pedal position information (APS%) of vehicle 10), data collected through the IMU (Inertial Measurement Unit) (e.g., vehicle 10's speed, acceleration, tilt angle, etc.), data collected through GPS (e.g., location information and altitude data), and data collected through other sensors (e.g., road environment information (slope, obstacles, etc.) collected through LiDAR / camera, presence and intensity of brake operation collected through brake sensors, etc.).

[0083] Furthermore, the driving logging data may include sensor data collected in real time from the vehicle 10 (e.g., real-time point cloud data collected via a LiDAR sensor) and positional result data obtained through precise map data matching corresponding to the location where the vehicle 10 is driving.

[0084] In step S120, the computing device 100 can generate training data using the driving logging data acquired in step S110.

[0085] In various embodiments, the computing device 100 can generate training data for training a first vehicle longitudinal control model.

[0086] Here, the first vehicle longitudinal control model is a model that derives result data for longitudinal control of the vehicle 10, and as shown in Figure 5, the first vehicle longitudinal control model is a model that derives result data for longitudinal control of the vehicle 10 (a dsr Since this model derives acceleration pedal position information (APS(%)) from vehicle state information as input, the training data may include target acceleration, vehicle state information, and acceleration pedal position information.

[0087] Here, longitudinal vehicle control can be a concept that includes not only acceleration control but also deceleration control.

[0088] When a driver directly operates the acceleration and brake pedals, they are used mutually exclusively, such as not using the acceleration pedal when using the brake pedal and not using the brake pedal when using the acceleration pedal. In contrast, when an autonomous driving system controls the vehicle 10, there are cases where the acceleration and brake pedals are operated simultaneously depending on the situation. Therefore, an acceleration model for controlling the vehicle's acceleration and a deceleration model for controlling its deceleration may be designed separately.

[0089] Considering these points, the training data for training the first vehicle longitudinal control model may include not only acceleration pedal position information, i.e., acceleration pedal position information for designing the acceleration model, but also brake pedal (deceleration pedal) position information for designing the deceleration control. However, it is not limited to this.

[0090] More specifically, first, the computing device 100 can obtain the target acceleration from the driving logging data.

[0091] In vehicle control systems, the engine and powertrain mechanisms (gears, shafts, tire inertia, etc.) have a characteristic that even when an input for acceleration (e.g., changing the position of the acceleration pedal) is made, the engine response is delayed due to the effects of mechanical inertia, the operation of the engine throttle and powertrain mechanisms, and turbo lag.

[0092] In other words, even if the acceleration pedal is operated to achieve the target acceleration at a specific point in time, such a response delay causes the acceleration of the vehicle 10 to be delayed and unable to immediately reach the target acceleration. As a result, the acceleration measured at a specific point in time may not match the target acceleration at that point in time.

[0093] Taking such response delays into account, the computing device 100 can determine the target acceleration for a specific time point by using the acceleration measured at a future time point as a reference point.

[0094] For example, the computing device 100 can acquire the acceleration measured from the vehicle 10 at a second time point, which is a predetermined time after the first time point, as the target acceleration at the first time point.

[0095] As another example, the computing device 100 can obtain the average value of accelerations measured from the vehicle 10 over a predetermined period of time after a specific point in time as the target acceleration at that specific point in time.

[0096] Next, the computing device 100 can acquire vehicle status information from the driving logging data.

[0097] Here, the vehicle status information may, but is not limited to, information regarding the longitudinal speed of the vehicle 10, engine RPM, and gear status.

[0098] In various embodiments, the computing device 100 can acquire vehicle status information such as the tilt information of the vehicle 10 (e.g., pitch angle) and the amount of change in the tilt of the vehicle 10 (e.g., change in pitch angle).

[0099] Even if the position of the acceleration pedal of vehicle 10 is the same, the acceleration of vehicle 10 will differ when it is traveling on flat ground compared to when it is traveling on an inclined section. By including the inclination information of vehicle 10 in the training data for learning the first vehicle longitudinal control model, it is possible to output result data that takes into account the inclination of the environment in which vehicle 10 is currently traveling, in the process of deriving result data for longitudinal control of vehicle 10 through the first vehicle longitudinal control model.

[0100] Furthermore, although the acceleration pedal position information for vehicle 10 was adjusted based on the driving logging data for vehicle 10, it is necessary to distinguish whether such adjustment of acceleration pedal position information is for adjusting the target acceleration or for maintaining the target acceleration even in inclined sections present on the driving route.

[0101] Taking these points into consideration, not only the tilt information of the vehicle 10 but also the amount of change in the tilt of the vehicle 10 can be included in the training data for learning the first vehicle longitudinal control model.

[0102] Such tilt information and tilt change information for vehicle 10 can be obtained based on driving logging data for vehicle 10, such as by extracting the vehicle pitch angle included in the driving logging data, or by calculating the difference in vehicle pitch angles at different points in time to obtain tilt change information. However, it is not limited to this, and the gradient and gradient change amount of the location where vehicle 10 is driving can be obtained based on pre-generated precision map data, and based on this, tilt information and tilt change information for vehicle 10 can be obtained.

[0103] Next, the computing device 100 can obtain acceleration pedal position information (APS(%)) from the driving logging data.

[0104] Here, the acceleration pedal position information indicates the degree to which the acceleration pedal is pressed, and may be expressed as a percentage. For example, if the state of not fully pressing the acceleration pedal is set to 0%, and the state of fully pressing it (maximum output) is set to 100%, the position of the acceleration pedal may be shown as a percentage based on this reference.

[0105] Typically, in an electronic throttle control (ETC) or drive-by-wire system, an electrical signal is generated in response to the amount the accelerator pedal is pressed, and the acceleration of the vehicle 10 is adjusted by adjusting the fuel injection throttle based on this electrical signal. Therefore, the amount the pedal is pressed can be estimated based on the electrical signal used to determine the amount of adjustment of the fuel injection throttle and / or the amount of adjustment of the fuel injection throttle, and through this, position information of the accelerator pedal can be obtained.

[0106] Next, the computing device 100 can generate learning data that includes the target acceleration and the acquired vehicle state information as input data, and acceleration pedal position information as ground truth data.

[0107] In various embodiments, the computing device 100 can preprocess the driving logging data (e.g., Figure 4) and generate training data based on the preprocessed driving logging data.

[0108] Here, preprocessing may or may not be limited to data scaling (e.g., standard scaling) which adjusts data values ​​to a certain range, and outlier data removal which identifies and removes outliers that deviate from the average pattern.

[0109] In step S130, the computing device 100 uses the learning data generated in step S120 to train a first vehicle longitudinal control model, thereby generating a first vehicle longitudinal control model that outputs acceleration pedal position information (APS(%)) using target acceleration and vehicle state information as input data.

[0110] Here, a first vehicle longitudinal control model (e.g., a neural network) consists of one or more network functions, and one or more network functions may consist of a set of interconnected computational units that can generally be referred to as “nodes.” Such “nodes” may also be referred to as “neurons.” One or more network functions consist of at least one or more nodes. The nodes (or neurons) that make up one or more network functions may be interconnected by one or more “links.”

[0111] In the first vehicle longitudinal control model, one or more nodes connected by links can form relative input node and output node relationships. The concepts of input and output nodes are relative; any node that is an output node to one node can be an input node to another node, and vice versa. As mentioned above, input node vs. output node relationships can be generated around links. One input node can be connected to one or more output nodes via links, and vice versa.

[0112] In an input-output node relationship connected through a single link, the output node's value can be determined based on the data input to the input node. Here, the nodes connecting the input and output nodes can have weights. These weights may be variable and can be varied by the user or algorithm to enable the first vehicle longitudinal control model to perform a desired function. For example, if one or more input nodes are interconnected to a single output node by their respective links, the output node's value can be determined based on the values ​​input to the input nodes connected to the output node and the weights set for the links corresponding to each input node.

[0113] As mentioned above, the first vehicle longitudinal control model consists of one or more nodes interconnected through one or more links, forming input node-output node relationships within the first vehicle longitudinal control model. The characteristics of the first vehicle longitudinal control model can be determined by the number of nodes and links within the first vehicle longitudinal control model, the correlation between nodes and links, and the weight values ​​assigned to each link. For example, if there are two first vehicle longitudinal control models with the same number of nodes and links but different weight values ​​between links, the two first vehicle longitudinal control models can be recognized as different from each other.

[0114] Some of the nodes constituting the first vehicle longitudinal control model can form a layer based on their distance from the initial input node. For example, a set of nodes that are n in distance from the initial input node can form an n-layer. The distance from the initial input node can be defined by the minimum number of links that must be traversed to reach that node. However, such a layer definition is arbitrary for illustrative purposes, and the order of layers within the first vehicle longitudinal control model can be defined in ways other than those described above. For example, the layer of nodes may be defined by their distance from the final output node.

[0115] The initial input node may mean one or more nodes in the first vehicle longitudinal control model to which data is directly input without going through links in relation to other nodes. Alternatively, it may mean a node in the first vehicle longitudinal control model network that does not have other input nodes connected to links in relation to links between nodes. Similarly, the final output node may mean one or more nodes in the first vehicle longitudinal control model that do not have output nodes in relation to other nodes. Furthermore, a hidden node may mean a node that constitutes the first vehicle longitudinal control model, rather than the initial input node or the final output node. In one embodiment of this disclosure, the first vehicle longitudinal control model may have more nodes in the input layer than in the hidden layer which is closer to the output layer, and the number of nodes may decrease as one progresses from the input layer to the hidden layer.

[0116] The first vehicle longitudinal control model may include one or more hidden layers. The hidden nodes of a hidden layer can take the output of the previous layer and the output of surrounding hidden nodes as inputs. The number of hidden nodes in each hidden layer may be the same or different. The number of nodes in the input layer may be determined based on the number of data fields in the input data, and may be the same as or different from the number of hidden nodes. The input data input to the input layer may be processed by the hidden nodes of the hidden layers and output by the fully connected layer (FCL), which is the output layer.

[0117] In various embodiments, the first vehicle longitudinal control model may be a deep learning model (e.g., Figure 6).

[0118] A deep learning model (e.g., a deep neural network (DNN)) can refer to a first-generation longitudinal control model that includes multiple hidden layers in addition to the input and output layers. By using deep neural networks, it is possible to understand the latent structures of data. That is, it is possible to understand the latent structures of photographs, texts, videos, audio, and music (for example, what objects are in a photograph, what is the content and emotion of the text, what is the content and emotion of the audio, etc.).

[0119] Deep neural networks include, but are not limited to, convolutional neural networks (CNNs), recurrent neural networks (RNNs), autoencoders, Generative Adversarial Networks (GANs), restricted Boltzmann machines (RBMs), deep belief networks (DBNs), Q networks, U networks, and Siam networks.

[0120] In various embodiments, the network function may include an autoencoder, which can be a type of artificial neural network that outputs output data similar to the input data.

[0121] An autoencoder may include at least one hidden layer, and an odd number of hidden layers may be placed between the input and output layers. The number of nodes in each layer may be reduced from the number of nodes in the input layer to an intermediate layer called a bottleneck layer (encoding), and then expanded from the bottleneck layer to the output layer (symmetrically with the input layer) through a reduction and symmetrical process. The nodes of the dimensionality reduction layer and the dimensionality restoration layer may or may not be symmetric. Furthermore, the autoencoder can perform nonlinear dimensionality reduction. The number of input and output layers can correspond to the number of sensors remaining after preprocessing of the input data. In an autoencoder structure, the number of nodes in the hidden layers included in the encoder may decrease as they move away from the input layer. The number of nodes in the bottleneck layer (the layer with the fewest nodes located between the encoder and decoder) may be maintained above a certain number (e.g., more than half the number of nodes in the input layer) because if it is excessively small, a sufficient amount of information may not be transmitted.

[0122] In various embodiments, the first vehicle longitudinal control model may be an LSTM (Long Short-Term Memory) model (e.g., Figure 7).

[0123] LSTM models are a type of cyclic neural network (RNN) and are deep learning models designed to process data where time dependence is important. While RNNs excel at processing time-series data, they suffer from gradient vanishing, where past information is forgotten when learning long sequences. LSTMs address this by managing both long-term and short-term memory to retain important information and forget unnecessary information. As mentioned earlier, in vehicle control systems, response delays can occur due to the engine and powertrain mechanisms (gears, shafts, tire inertia, etc.). Therefore, simply considering data at a specific point in time is insufficient, and it is necessary to analyze past and / or future data together. Considering these points, LSTM models, which excel at processing time-series data, can be utilized as the first vehicle longitudinal control model.

[0124] In other words, LSTMs are models designed to solve long-term dependency problems, effectively memorizing past input information and generating current control instructions based on it. Such characteristics can be very advantageous in handling temporal nonlinearities, such as engine response delays.

[0125] Furthermore, unlike existing control techniques (e.g., PID control, MPC control), which rely on fixed models, LSTM-based control models differ significantly in that they are dynamically learned through data. Through this, LSTMs can flexibly respond to a variety of driving conditions that are difficult for fixed models to handle.

[0126] Furthermore, LSTM-based control models offer significant advantages in terms of real-time control performance. Unlike existing model predictive controllers (MPC controllers), LSTMs do not need to solve complex optimization problems in real time. This means that the computational cost is lower when generating control commands in real time, providing the advantage of being able to quickly adapt to sudden changes in vehicle speed and unpredictable environmental changes.

[0127] In one embodiment, the LSTM-based first vehicle longitudinal control model according to the present invention can receive state variables as input, process time-series characteristics, and operate to control the accelerator pedal based on these characteristics. As an example, the first vehicle longitudinal control model may consist of an input encoder, an LSTM cell, and an output layer.

[0128] Firstly, the input encoder receives vehicle data as input and converts it into a dimension suitable for the LSTM cell. The vehicle data includes various vehicle state information such as engine speed and brake pedal position, and this data is converted into an encoding vector and transmitted to the LSTM cell.

[0129] Secondly, the LSTM cell processes time-series data by receiving an encoded input vector provided by the input encoder. This cell can be a core module that stores past driving information and predicts the vehicle's future state based on this. Through the LSTM cell, it is possible to learn and predict changes in the vehicle's state over time.

[0130] Finally, the output layer receives the data output from the LSTM cells and generates the final control signals. This layer post-processes the output of the LSTM cells to predict the vehicle's state, such as speed and acceleration. Through the output layer, the model generates the actual control signals based on the predicted vehicle state, thereby enabling longitudinal control of the vehicle.

[0131] As an example, an LSTM-based first vehicle longitudinal control model may be trained to minimize the error between the target accelerator pedal position and the estimated accelerator pedal position (e.g., the measured accelerator pedal position) using a Mean Squared Error (MSE) loss function.

[0132] Figures 8 and 9 illustrate the results of training the first vehicle longitudinal control model using the method described above. Referring to Figures 8 and 9, it can be seen that the first vehicle longitudinal control model accurately predicts the acceleration pedal position information, and that training is performed while reducing both the training loss and the validation loss.

[0133] Furthermore, Figure 10 illustrates the experimental results of the first vehicle longitudinal control model learned by the method described above. Referring to Figure 10, it can be seen that the vehicle 10 is controlled as intended by the result data derived through the first vehicle longitudinal control model.

[0134] Figure 11 is a flowchart of a method for generating a second vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of an autonomous vehicle in various embodiments.

[0135] Referring to Figure 11, at step S210, the computing device 100 can acquire driving logging data for the vehicle 10.

[0136] Here, the method for acquiring driving logging data for vehicle 10 and the driving logging data acquired therethrough may be the same as, but not limited to, the method for acquiring driving logging data at step S110 in Figure 3 and the driving logging data acquired therethrough.

[0137] In step S220, the computing device 100 can generate training data using the driving logging data acquired in step S210.

[0138] In various embodiments, the computing device 100 can generate training data for training a second vehicle longitudinal control model.

[0139] Here, the second vehicle longitudinal control model, like the first vehicle longitudinal control model, is a model that derives result data for vehicle longitudinal control, and as shown in Figure 12, the second vehicle longitudinal control model determines the target acceleration (a) of the vehicle 10. dsr Since this model derives the optimal combination of parameter functions to constitute a longitudinal vehicle model using vehicle state information as input, the training data may include target acceleration, vehicle state information, and estimated acceleration.

[0140] More specifically, the computing device 100 can first acquire target acceleration and vehicle status information from the driving logging data.

[0141] Here, the method for obtaining target acceleration and vehicle state information from driving logging data may be implemented in the same and / or similar form as the method in step S120 of Figure 3, but is not limited thereto.

[0142] Thereafter, the computing device 100 can acquire the target acceleration and the estimated acceleration corresponding to the vehicle status information.

[0143] In various embodiments, the computing device 100 can calculate estimated acceleration through a vehicle dynamics model based on target acceleration and vehicle state information. However, the estimated acceleration may, but is not limited to, the actual acceleration measured from the vehicle 10.

[0144] Here, the vehicle dynamics model can be expressed as shown in mathematical equation 1 below.

[0145]

number

[0146] Here, TIFF2026106386000018.tif66 is engine output (driving force), TIFF2026106386000019.tif610 is a rolling resistance, TIFF2026106386000020.tif1013 is air resistance, TIFF2026106386000021.tif99 is pitching resistance and TIFF2026106386000022.tif99 could be an inertial drag.

[0147] Next, the computing device 100 can generate training data that includes target acceleration and vehicle state information as input data, and estimated acceleration as ground truth data.

[0148] In step S230, the computing device 100 uses the learning data generated in step S220 to train a second vehicle longitudinal control model, thereby generating a second vehicle longitudinal control model that takes target acceleration and vehicle state information as input data to derive the optimal combination of parameter functions that constitute the vehicle longitudinal model derived from the vehicle dynamics model.

[0149] Here, the configuration and structure of the second vehicle longitudinal control model may be, but are not limited to, the same as those of the first vehicle longitudinal control model.

[0150] Furthermore, the vehicle longitudinal model derived from the vehicle dynamics model can be a mathematical representation of the vehicle's longitudinal motion. For example, the vehicle longitudinal model can be expressed as shown in mathematical equation 2 below.

[0151] <Mathematical formula 2>

number

[0152] Here, TIFF2026106386000024.tif54 is a vehicle longitudinal model. TIFF2026106386000025.tif59 is a gear scaling coefficient function based on the gear condition. TIFF2026106386000026.tif65 contains static parameters related to gears. TIFF2026106386000027.tif54 is gear status, as mentioned above. TIFF2026106386000028.tif711 is a gear parameter function based on speed. TIFF2026106386000029.tif43 is speed, TIFF2026106386000030.tif515 contains acceleration pedal position values ​​and TIFF2026106386000031.tif725 may be a DC gain function that changes depending on the accelerator pedal position and gear state.

[0153] Furthermore, here is the vehicle longitudinal model. TIFF2026106386000032.tif54 may physically represent the longitudinal acceleration of vehicle 10, and a vehicle longitudinal model can be used to calculate longitudinal acceleration by considering various factors acting on the vehicle during travel.

[0154] The optimal combination of parameter functions to be derived based on the second vehicle longitudinal control model is the parameter function relating to the gear scaling coefficient depending on the gear state (e.g., the function relating to the gear scaling coefficient depending on the gear state ( TIFF2026106386000033.tif59)) (hereinafter referred to as the "first parameter function"), gear-related parameter functions (e.g., gear-related static parameters ( TIFF2026106386000034.tif65) and gear parameter function based on speed ( TIFF2026106386000035.tif711)) (hereinafter referred to as the "second parameter function"), a parameter function related to the DC gain which changes depending on the acceleration pedal position and gear state ( TIFF2026106386000036.tif725) (hereinafter referred to as the "third-parameter function") may be included, but is not limited to.

[0155] In other words, the computing device 100 includes target acceleration and vehicle state information as input data, and uses training data including estimated acceleration as ground truth data to train a second vehicle longitudinal control model. This generates a second vehicle longitudinal control model that can derive the optimal combination of parameter functions that minimizes the error between target acceleration and estimated acceleration in a specific vehicle state.

[0156] In various embodiments, the computing device 100 can generate multiple neural network models as unit models for each of the multiple parameter functions using training data that includes target acceleration and vehicle state information as input data and estimated acceleration as ground truth data, and can generate a single vehicle longitudinal control model that includes multiple neural network models. That is, after individually designing unit neural network models for each of the multiple parameter functions included in the vehicle longitudinal control model, the vehicle longitudinal control model that includes the multiple neural network models can be trained as a single model using the training data to generate the vehicle longitudinal control model.

[0157] Each parameter function that makes up a vehicle longitudinal model depends on different variables. However, if a single model is designed to derive various parameter functions without considering these differences, problems may arise where the parameter functions are affected by other variables. For example, even though the first parameter function actually depends on the gear state, if it is designed as part of a single model, problems may arise where it is affected by the vehicle speed 10 or engine RPM.

[0158] Taking these points into consideration, the computing device 100 according to the present invention can prevent the parameter function from being affected by other variables by individually designing a neural network model as a unit model for each parameter function. Furthermore, in order to train the neural network model to accurately represent the actual acceleration of the vehicle 10 based on the vehicle 10's driving logging data, each neural network model can be trained together as a single neural network.

[0159] Here, a neural network model designed individually for each parameter function may be a model that has learned only the correlation between each parameter function and the information on which each parameter function depends.

[0160] For example, a first neural network model designed for a first parameter function may be a neural network model that has learned only the correlation between the first parameter function and the gear state. Similarly, a second neural network model designed for a second parameter function may be a neural network model that has learned only the correlation between the second parameter function and the velocity and gear state. Furthermore, a third neural network model designed for a third parameter function may be a neural network model that has learned only the correlation between the third parameter function and the acceleration pedal position information and the gear state.

[0161] The method for generating a vehicle longitudinal control model for the artificial intelligence platform for longitudinal control of autonomous vehicles described above has been explained with reference to the flowchart illustrated in the drawings. For the sake of brevity, the method for generating a vehicle longitudinal control model for the artificial intelligence platform for longitudinal control of autonomous vehicles has been illustrated and explained in a series of blocks, but this disclosure is not limited to the order of the blocks, and some blocks may be performed in a different order or simultaneously than those illustrated and described herein. Furthermore, new blocks not described herein and in the drawings may be added, or some blocks may be deleted or modified.

[0162] While embodiments of this disclosure have been described above with reference to the attached drawings, a person ordinary in the art to which this disclosure belongs will understand that this disclosure may be implemented in other specific forms without altering its technical idea or essential features. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive. [Explanation of symbols]

[0163] 100: Computing device 200: User terminal 300: External Server 400: Network

Claims

1. In a method for generating a vehicle longitudinal control model for an artificial intelligence-based system for the longitudinal control of an autonomous vehicle, performed by a computing device, The stage of acquiring driving logging data for the vehicle; A step of generating training data using the aforementioned acquired driving logging data; and A method for generating a vehicle longitudinal control model for longitudinal control of an autonomous vehicle, comprising the step of generating a vehicle longitudinal control model that derives result data for longitudinal control of the vehicle using the generated training data.

2. The step of generating the aforementioned training data is: The step of obtaining the target acceleration from the aforementioned acquired driving logging data; The step of obtaining vehicle status information from the aforementioned acquired driving logging data; A step of obtaining acceleration pedal position information from the aforementioned acquired driving logging data; and A method for generating a vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of an autonomous vehicle, according to claim 1, comprising the step of generating learning data which includes the acquired target acceleration and the acquired vehicle state information as input data and the acquired acceleration pedal position information as correct answer data.

3. The step of achieving the aforementioned target acceleration is: A method for generating a vehicle longitudinal control model for an artificial intelligence-based system for longitudinal control of an autonomous vehicle, according to claim 2, further comprising the step of obtaining the acceleration measured from the vehicle at a second time point, which is a predetermined time after the first time point, as the target acceleration at the first time point.

4. The step of achieving the aforementioned target acceleration is: A method for generating a vehicle longitudinal control model for an artificial intelligence-based system for longitudinal control of an autonomous vehicle, according to claim 2, further comprising the step of obtaining the average value of accelerations measured from the vehicle for a predetermined period of time after a specific point in time as the target acceleration at the specific point in time.

5. The step of acquiring the aforementioned vehicle status information is: A method for generating a vehicle longitudinal control model for an artificial intelligence-based vehicle longitudinal control according to claim 2, comprising the step of obtaining information regarding the vehicle's longitudinal speed, engine RPM, and gear status from the acquired driving logging data as vehicle status information.

6. The step of acquiring the aforementioned vehicle status information is: A method for generating a vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of an autonomous vehicle, according to claim 2, comprising the step of acquiring at least one of the following information from the acquired driving logging data as vehicle state information: vehicle tilt information and information on the amount of change in the vehicle tilt.

7. The step of generating the aforementioned training data is: A step in which the vehicle acquires information about the driving environment while it is in motion, based on pre-generated precision map data; and This includes the step of generating training data using the acquired driving environment information and the acquired driving logging data, The aforementioned acquired driving environment information is A method for generating a vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of an autonomous vehicle, according to claim 1, wherein the method includes at least one of the inclination of the location where the vehicle is traveling and the amount of change in inclination.

8. The step of generating training data using the acquired driving environment information and the acquired driving logging data is as follows: A step of acquiring at least one piece of information as vehicle state information, consisting of the vehicle's tilt information and the amount of tilt change information, based on the inclination and the amount of change in inclination of the location where the vehicle is traveling; and A method for generating a vehicle longitudinal control model for an artificial intelligence-based vehicle longitudinal control of an autonomous vehicle, according to claim 7, comprising the step of generating training data using the at least one piece of information acquired and the acquired driving logging data.

9. The generated vehicle longitudinal control model is, A method for generating a vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of an autonomous vehicle, as described in claim 1, which is an LSTM (Long Short-Term Memory) model.

10. The step of generating the aforementioned training data is: The step of obtaining the target acceleration from the aforementioned acquired driving logging data; The step of obtaining vehicle status information from the aforementioned acquired driving logging data; A step of extracting the estimated acceleration corresponding to the acquired target acceleration and the acquired vehicle state information; and The process includes generating learning data that includes the acquired target acceleration and the acquired vehicle state information as input data, and the extracted estimated acceleration as ground truth data, The step of generating the vehicle longitudinal control model is: A method for generating a vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of an autonomous vehicle, according to claim 1, comprising the step of generating a vehicle longitudinal control model that derives the optimal combination of parameter functions constituting the vehicle longitudinal model derived from the vehicle dynamics model by training using the generated training data.

11. The step of generating a vehicle longitudinal control model that derives the optimal combination of parameter functions constituting the aforementioned vehicle longitudinal model is: If the vehicle longitudinal model includes multiple parameter functions, the step of generating multiple neural network models as unit models for each of the multiple parameter functions; and A method for generating a vehicle longitudinal control model on an artificial intelligence platform for longitudinal control of an autonomous vehicle, according to claim 10, further comprising the step of generating a single vehicle longitudinal control model including the plurality of neural network models generated.

12. The aforementioned vehicle dynamics model is, A method for generating a vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of an autonomous vehicle, as described in claim 10, which can be expressed as shown in mathematical formula 1 below. <Mathematical formula 1> [Math 1] Here, the above 【number】 is the engine output, the above 【number】 is rolling resistance, the aforementioned 【number】 This is air resistance, the aforementioned 【number】 The resistance to pitching and the aforementioned 【number】 This is inertial resistance.

13. The aforementioned vehicle longitudinal model is, A method for generating a vehicle longitudinal control model for an artificial intelligence platform for longitudinal control of an autonomous vehicle, as described in claim 10, which can be expressed as shown in mathematical formula 2 below. <Mathematical formula 2> [Math 2] Here, the above 【number】 This is a vehicle longitudinal model, the above 【number】 This is a parameter function relating to the gear scaling coefficient depending on the gear condition, 【number】 These are static parameters related to the gear, the aforementioned 【number】 The gear state, the aforementioned 【number】 The gear parameter function depends on the speed, 【number】 is speed, the above 【number】 The acceleration pedal position value and the above 【number】 This is a DC gain function that changes depending on the accelerator pedal position and gear state.

14. The optimal combination of parameter functions derived above is: A method for generating a vehicle longitudinal control model for an artificial intelligence-based system for longitudinal control of an autonomous vehicle, according to claim 10, comprising a parameter function relating to a gear scaling coefficient depending on the gear state, a parameter function related to the gear, and a parameter function relating to the DC gain which changes depending on the acceleration pedal position value and the gear state.

15. The step of generating the aforementioned training data is: The step of preprocessing the acquired driving logging data; and A method for generating a vehicle longitudinal control model for an artificial intelligence base for the longitudinal control of an autonomous vehicle, according to claim 1, further comprising the step of generating training data using the pre-processed driving logging data.

16. Processor; Network interface; memory; and The memory is loaded and the computer program is executed by the processor, The aforementioned computer program, Instructions for acquiring vehicle driving logging data; Instructions for generating training data using the aforementioned acquired driving logging data; and A computing device that performs a method for generating a vehicle longitudinal control model for the longitudinal control of an autonomous vehicle, based on artificial intelligence, including instructions for generating a vehicle longitudinal control model that derives result data for vehicle longitudinal control using the generated training data.

17. Combined with computing devices, The stage of acquiring driving logging data for the vehicle; A step of generating training data using the aforementioned acquired driving logging data; and A computer program stored on a recording medium readable by a computing device for executing an artificial intelligence-based method for generating a vehicle longitudinal control model for the longitudinal control of an autonomous vehicle, which includes the step of generating a vehicle longitudinal control model that derives result data for vehicle longitudinal control using the generated training data.

Images