Vehicle control apparatus and method thereof

Abstract

The invention relates to a device for controlling the travel of a vehicle. The apparatus includes: a memory storing at least one instruction; and a processor operably coupled with the memory. The at least one instruction, when executed by the processor, is configured to cause the apparatus to: obtain a data set containing a plurality of frames for travel control of the vehicle; classifying the data set into a plurality of bundles; and dividing the plurality of beams into training data and evaluation data. And executing a precision test based on the training data and the evaluation data meeting a first condition. And training the artificial intelligence model or evaluating the performance of the artificial intelligence model based on the precision test meeting a second condition. The apparatus may output a signal based on the trained artificial intelligence model or the evaluated performance of the artificial intelligence model, and control driving of the vehicle based on the signal.

Description

Technical Field

[0001] This disclosure relates to a vehicle control device and method, and more particularly to a technique for classifying a dataset for vehicle driving control. Background Technology

[0002] The descriptions in this background section are intended only to enhance the understanding of the background of this disclosure and should not be construed as an admission that they correspond to prior art known to those skilled in the art.

[0003] With the development of autonomous driving control technology and / or semi-autonomous driving control (or cruise driving) technology, the stable driving technology of this vehicle can gradually become more mature. For example, when identifying situations requiring deceleration, lateral movement, or lane changes during driving control of this vehicle, it is necessary to develop various algorithms for executing driving strategies determined based on the various conditions of the identified situations.

[0004] On the other hand, in order to achieve autonomous driving control, it is necessary to accurately and quickly process the sensor data obtained through sensors. In particular, it is necessary to develop technologies that use various artificial intelligence models (e.g., deep learning models) to process sensor data to detect and identify the vehicle's surrounding environment (e.g., external objects).

[0005] Furthermore, if an artificial intelligence model is used, the data can be divided into two categories. For example, training data used to train the artificial intelligence model and evaluation data used for evaluation can be data that is divided into two categories for the purpose of using the artificial intelligence model.

[0006] However, if the dataset is simply divided according to a predetermined ratio, specific parameters will only be concentrated in some data, resulting in reduced accuracy. Therefore, an adaptive benchmark is needed to divide the dataset into evaluation and training data. Summary of the Invention

[0007] According to this disclosure, an apparatus for controlling the movement of a vehicle may include a memory storing at least one instruction and a processor operatively coupled to the memory, wherein the at least one instruction, when executed by the processor, is configured to cause the apparatus to: obtain a dataset comprising multiple frames for driving control of the vehicle; classify the dataset into multiple bundles; divide the multiple bundles into training data and evaluation data; perform an accuracy test on the training data and evaluation data based on a first condition satisfying the training data and evaluation data; train an artificial intelligence model or evaluate the performance of an artificial intelligence model based on a second condition satisfying the accuracy test; output a signal based on the performance of the trained artificial intelligence model or the evaluated artificial intelligence model; and control the movement of the vehicle based on the signal.

[0008] The device may also include sensors, wherein at least one instruction, when executed by the processor, is configured to cause the device to: input sensor data obtained using the sensors into an artificial intelligence model to detect objects present outside the vehicle during driving control.

[0009] Multiple frames may include at least one of the following: the category of the external object, the location of the external object, the size of the external object, the acquisition period, surrounding traffic information, global positioning system (GPS) information, and weather information, wherein the acquisition period corresponds to the duration for which data is collected for the frames within the bundle.

[0010] At least one instruction, when executed by a processor, is configured to cause the device to: identify a first number of specific objects contained in a first frame, wherein the first frame is divided into a first bundle of training data in a plurality of bundles; identify a second number of specific objects contained in a second frame, wherein the second frame is divided into a second bundle of evaluation data in a plurality of bundles; and determine that a first condition is satisfied based on the ratio between the first number and the second number being within a certain error range relative to a predefined ratio.

[0011] At least one instruction, when executed by the processor, is configured to cause the device to: acquire multiple frames using sensors; and classify the dataset into multiple bundles based on the acquisition time of each of the multiple frames.

[0012] At least one instruction, when executed by the processor, is configured to cause the device to: shorten the baseline time for classifying the dataset into multiple bundles based on the failure to meet the first condition, and classify the dataset into multiple bundles again.

[0013] At least one instruction, when executed by the processor, is configured to cause the device to: identify multiple partitioning results that divide multiple bundles into training data and evaluation data; and, based on the fact that none of the multiple partitioning results satisfy the first condition, reclassify the dataset into different multiple bundles.

[0014] At least one instruction, when executed by the processor, is configured to cause the device to: determine that a second condition is satisfied based on the difference between a first inclusion percentage and a second inclusion percentage being less than or equal to a specific percentage, wherein the first inclusion percentage corresponds to the proportion of a first frame containing each external object, wherein the first frame is divided into a first bundle of training data in a plurality of bundles, and the second inclusion percentage corresponds to the proportion of a second frame containing each external object, wherein the second frame is divided into a second bundle of evaluation data in a plurality of bundles.

[0015] At least one instruction, when executed by the processor, is configured to cause the device to: determine that a second condition is satisfied based on the difference between a first standard deviation and a second standard deviation being less than or equal to a specific value, wherein the first standard deviation corresponds to the standard deviation of the probability distribution of each external object contained in the first frame, and the second standard deviation corresponds to the standard deviation of the probability distribution of each external object contained in the second frame.

[0016] At least one instruction, when executed by the processor, is configured to cause the device to: further determine whether a second condition is satisfied based on at least one of the following: weather information for each of the training data and the evaluation data, traffic congestion for each of the training data and the evaluation data, global positioning system (GPS) information for each of the training data and the evaluation data, and the number of vehicles per frame for each of the training data and the evaluation data.

[0017] According to this disclosure, a method performed by a device for controlling the driving of a vehicle may include: obtaining a dataset containing multiple frames for driving control of the vehicle; classifying the dataset into multiple bundles; dividing the multiple bundles into training data and evaluation data; performing an accuracy test on the training data and evaluation data based on a first condition being met; training an artificial intelligence model or evaluating the performance of the artificial intelligence model based on a second condition being met by the accuracy test; outputting a signal based on the performance of the trained artificial intelligence model or the evaluated artificial intelligence model; and controlling the driving of the vehicle based on the signal.

[0018] The method may also include: inputting sensor data obtained using the vehicle's sensors into an artificial intelligence model to detect objects present outside the vehicle during driving control.

[0019] Multiple frames may include at least one of the following: the category of the external object, the location of the external object, the size of the external object, the acquisition period, surrounding traffic information, global positioning system (GPS) information, and weather information, wherein the acquisition period corresponds to the duration for which data is collected for the frames within the bundle.

[0020] The method may further include: identifying a first number of specific objects contained in a first frame, wherein the first frame is divided into a first bundle of training data in a plurality of bundles; identifying a second number of specific objects contained in a second frame, wherein the second frame is divided into a second bundle of evaluation data in a plurality of bundles; and determining that a first condition is satisfied based on the ratio between the first number and the second number being within a certain error range relative to a predefined ratio.

[0021] The method may also include: acquiring multiple frames using a sensor; and classifying the dataset into multiple bundles based on the acquisition time of each frame in the multiple frames.

[0022] The method may further include: shortening the baseline time for classifying the dataset into multiple bundles based on the failure to meet the first condition, and classifying the dataset into multiple bundles again.

[0023] The method may further include: identifying multiple partitioning results that divide multiple bundles into training data and evaluation data; and classifying the dataset into different multiple bundles again based on the fact that none of the multiple partitioning results satisfy the first condition.

[0024] The method may further include: determining a second condition based on the difference between a first inclusion percentage and a second inclusion percentage being less than or equal to a specific percentage, wherein the first inclusion percentage corresponds to the proportion of a first frame containing each external object, wherein the first frame is divided into a first bundle of training data in a plurality of bundles, and the second inclusion percentage corresponds to the proportion of a second frame containing each external object, wherein the second frame is divided into a second bundle of evaluation data in a plurality of bundles.

[0025] The method may further include: determining a second condition based on the difference between a first standard deviation and a second standard deviation being less than or equal to a specific value, wherein the first standard deviation corresponds to the standard deviation of the probability distribution of each external object contained in the first frame, and the second standard deviation corresponds to the standard deviation of the probability distribution of each external object contained in the second frame.

[0026] The method may further include determining whether a second condition is met based on at least one of the following: weather information for the training data and the evaluation data respectively, traffic congestion for the training data and the evaluation data respectively, global positioning system (GPS) information for the training data and the evaluation data respectively, and the number of vehicles per frame for the training data and the evaluation data respectively. Attached Figure Description

[0027] The above and other objects, features and advantages of this disclosure will become more apparent from the following detailed description taken in conjunction with the accompanying drawings:

[0028] Figure 1 An example of a component of a vehicle control device according to an example of this disclosure is shown;

[0029] Figure 2 An example is shown of the result of classifying multiple frames into multiple bundles in a vehicle control device according to an example of the present disclosure;

[0030] Figure 3 An example is shown of the result of dividing multiple bundles into training data or evaluation data in a vehicle control device according to an example of the present disclosure;

[0031] Figure 4Examples of multiple division results of multiple bundles in a vehicle control device according to the present disclosure are shown;

[0032] Figure 5 An example is shown of the number of objects contained in the training data and evaluation data according to the examples of this disclosure;

[0033] Figure 6A Examples of specific objects included in training and evaluation data are shown in a specific technique according to the present disclosure;

[0034] Figure 6B Examples of specific objects included in training and evaluation data are shown in a specific technique according to the present disclosure;

[0035] Figure 6C Examples of specific objects included in training and evaluation data are shown in a specific technique according to the present disclosure;

[0036] Figure 6D Examples of specific objects included in training and evaluation data are shown in a specific technique according to the present disclosure;

[0037] Figure 7 An example of a vehicle control method according to the present disclosure is shown; and

[0038] Figure 8 An example of a computational system for a vehicle control method is shown according to an example of this disclosure.

[0039] Regarding the description of the accompanying drawings, the same or similar reference numerals may be used for the same or similar parts. Detailed Implementation

[0040] The following describes some examples of this disclosure in detail with reference to the exemplary accompanying drawings. When adding reference numerals to components in the various drawings, it should be noted that the same components will be represented by the same reference numerals even if they appear in other drawings. Furthermore, to avoid unnecessarily obscuring the spirit of this disclosure, detailed descriptions of known features or functions will be omitted.

[0041] In describing the components of this disclosure, terms such as first, second, "A", "B", (a), (b), etc., may be used. Regardless of the order or priority of the corresponding elements, these terms are used only to distinguish one element from another and do not limit the corresponding element. Furthermore, unless otherwise defined, all terms used herein, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. Such terms, as defined in common dictionaries, should be interpreted as having the same meaning as in the context of the relevant field, and should not be interpreted as having an ideal or extra-normative meaning, unless expressly defined in this application as having such a meaning.

[0042] For the purposes of this application and claims, the exemplary phrases “at least one of the following: A; B; or C” or “at least one of A, B, or C” are used, which means “at least one A, or at least one B, or at least one C, or any combination of at least one A, at least one B, and at least one C.” Furthermore, exemplary phrases such as “A, B, and C,” “A, B, or C,” “at least one of A, B, and C,” and “at least one of A, B, or C” can herein represent each of the listed items or all possible combinations of the listed items. For example, “at least one of A or B” can mean (1) at least one A; (2) at least one B; or (3) at least one A and at least one B.

[0043] According to the Society of Automotive Engineers (SAE) standards, the automation levels of autonomous vehicles can be classified as follows: At Level 0 of Automated Driving, the SAE classification corresponds to "No Automation," meaning the automated driving system intervenes only temporarily in emergency situations (e.g., automatic emergency braking) and / or only provides warnings (e.g., blind spot warning, lane departure warning, etc.), and expects the driver to operate the vehicle. At Level 1 of Automated Driving, the SAE classification corresponds to "Driver Assistance," meaning the system performs some driving functions (e.g., steering, acceleration, braking, lane centering, adaptive cruise control, etc.) during normal driver operation, and expects the driver to determine the system's operating status and / or timing, perform other driving functions, and respond to (e.g., resolve) emergency situations. At Level 2 of Automated Driving, the SAE classification corresponds to "Partial Automation," meaning the system performs steering, acceleration, and / or braking under driver supervision, and expects the driver to determine the system's operating status and / or timing, perform other driving functions, and respond to (e.g., resolve) emergency situations. At Level 3 of automated driving, the SAE classification standard can correspond to "conditional automation," meaning the system drives the vehicle under limited conditions (e.g., performing driving functions such as steering, acceleration, and / or braking), but transfers driving control to the driver if the required conditions are not met. The driver is expected to determine the system's operating state and / or timing, and take over control in emergency situations, but not operate the vehicle (e.g., steering, acceleration, and / or braking) in other situations. At Level 4 of automated driving, the SAE classification standard can correspond to "high automation," meaning the system performs all driving functions, and the driver is expected to control the vehicle only in emergency situations. At Level 5 of automated driving, the SAE classification standard can correspond to "full automation," meaning the system performs all driving functions without any driver assistance in all situations, including emergency situations, and the driver is expected to perform no driving functions other than determining the system's operating state. While the SAE automated driving classification standard can be applied to this disclosure, other classification methods and / or algorithms may also be used in one or more configurations described herein.

[0044] One or more features related to autonomous driving control can be activated based on configured autonomous driving control settings (e.g., based on at least one of the following: autonomous driving classification, selection of vehicle autonomous driving level, etc.). Vehicle operation can be controlled based on one or more features described herein (e.g., features that classify a dataset into bundles and divide the bundles into training and evaluation data). Vehicle control can include various vehicle-related operational controls (e.g., autonomous driving control, sensor control, braking control, braking timing control, acceleration control, rate of change of acceleration control, warning timing control, pre-collision warning timing control, etc.).

[0045] For example, one or more auxiliary devices (e.g., engine braking, exhaust braking, hydraulic retarder, electric retarder, regenerative braking, etc.) can also be controlled based on one or more features described herein (e.g., features that classify datasets into bundles and divide bundles into training data and evaluation data).

[0046] For example, one or more communication devices (e.g., modems, network adapters, radio transceivers, antennas, etc., capable of communicating via one or more wired or wireless communication protocols, such as Ethernet, Wi-Fi, Near Field Communication (NFC), Bluetooth, Long Term Evolution (LTE), 5G New Radio (NR), Vehicle-to-Everything (V2X) etc.) can also be controlled based on one or more features described herein (e.g., features that classify datasets into bundles and divide bundles into training data and evaluation data).

[0047] Minimum Risk Maneuver (MRM) operations can also be controlled based on one or more features described herein (e.g., features that classify the dataset into bundles and divide the bundles into training and evaluation data). A minimum risk maneuver operation (e.g., a minimum risk maneuver, a minimum risk operation) can be a maneuver of the vehicle to minimize (e.g., reduce) the risk of collision with surrounding vehicles, thereby achieving a lower (e.g., minimum) risk state. A minimum risk maneuver can be an operation that can be activated during autonomous driving if the driver is unable to respond to an intervention request. During a minimum risk maneuver, one or more processors of the vehicle can control the vehicle's driving operations for a set time period.

[0048] For example, biased driving operations can also be controlled based on one or more features described herein (e.g., features that classify datasets into bundles and divide bundles into training and evaluation data). A driving control device can perform biased driving control. To perform biased driving, the driving control device can control the vehicle to travel within the lane by maintaining a lateral distance between the vehicle's center position and the lane center. For example, the driving control device can control the vehicle to remain within the lane but not in the lane center. The driving control device can identify or determine a target lateral distance for biased driving control. For example, the target lateral distance can include an intentionally adjusted lateral distance that the vehicle aims to maintain with respect to a reference point (e.g., the lane center or another vehicle) during maneuvers such as lane changes. Such adjustments can be made to improve the vehicle's stability, safety, and / or performance under different driving conditions. For example, during a lane change, the driving control system can adjust the lateral distance to maintain a safer distance from adjacent vehicles, taking into account factors such as vehicle speed, road conditions, and / or the presence of obstacles.

[0049] One or more sensors (e.g., IMU (Inertial Measurement Unit) sensors, cameras, LiDAR, radar, blind spot monitoring sensors, lane departure warning sensors, parking sensors, light sensors, rain sensors, traction control sensors, anti-lock braking system sensors, tire pressure monitoring sensors, seat belt sensors, airbag sensors, fuel sensors, emission sensors, throttle position sensors, inverters, converters, motor controllers, power distribution units, high-voltage cables and connectors, auxiliary power modules, charging interfaces, etc.) can also be used for control, for example, based on one or more features described herein (e.g., features that classify datasets into bundles and divide bundles into training data and evaluation data). The operational control of autonomous driving can include various driving controls of the vehicle by the vehicle control unit (e.g., acceleration, deceleration, steering control, gear shifting control, braking system control, traction control, stability control, cruise control, lane keeping assist control, collision avoidance system control, emergency braking assist control, traffic sign recognition control, adaptive headlight control, etc.).

[0050] The following will refer to Figures 1 to 8 Examples of this disclosure are described in detail.

[0051] Figure 1 An example of a component of a vehicle control device according to this disclosure is shown.

[0052] According to one example, the vehicle control unit 100 may include at least one of a memory 110, a processor 120, a sensor 130, and any combination thereof. Figure 1 The components of the vehicle control device 100 shown are merely exemplary, and the examples disclosed herein are not limited thereto. For example, the vehicle control device 100 may also include... Figure 1 Components not shown (e.g., at least one of an interface, communication device, display, drive device, and any combination thereof).

[0053] According to one example, memory 110 can store commands or data. For example, memory 110 can store one or more instructions that, when executed by processor 120, cause vehicle control unit 100 to perform various operations.

[0054] For example, memory 110 and processor 120 can be implemented as a single chipset. Processor 120 may include at least one of a communication processor and a modem.

[0055] For example, memory 110 can store the operating history of vehicle control device 100. For example, memory 110 can store datasets acquired by sensor 130. For example, memory 110 can store results divided into training data and evaluation data under the control of processor 120. Training data can be a subset of the dataset used to train machine learning models by enabling them to learn patterns and relationships. Evaluation data can be used to test the model's performance on unseen data to ensure that the model can generalize effectively. Training and evaluation data can be distinct and non-overlapping to avoid overfitting and ensure reliable evaluation metrics. Both training and evaluation data can have balanced and similar distributions to reflect the overall dataset and avoid biased or skewed results. To achieve a balanced and similar distribution, the dataset can be partitioned in a deliberate manner (e.g., ensuring that rare object types or specific environmental conditions exist in two subsets in comparable proportions) or in a systematic manner (e.g., avoiding imbalances or unexpected biases that may result from simple random partitioning) to ensure that the two subsets are distinct (e.g., there is no overlap between training and evaluation data to avoid overfitting), balanced (e.g., the two subsets have similar key distributional features, such as object categories, environmental conditions, or collection settings), and representative (e.g., each subset accurately reflects the diversity of the overall dataset, covering a wide range of situations to prevent bias).

[0056] According to one example, processor 120 may be operatively connected to memory 110 and / or sensor 130. For example, processor 120 may control the operation of memory 110 and / or sensor 130.

[0057] For example, processor 120 can obtain a dataset for driving control of the vehicle.

[0058] As an example, the dataset may include multiple frames used for driving control of the vehicle.

[0059] As an example, processor 120 can use sensor 130 to acquire a dataset. The multiple frames contained in the dataset can include various information, such as information about the vehicle's driving environment.

[0060] As an example, multiple frames can include information about external objects (e.g., another vehicle, person, object, building, road structure, etc.) such as the type of another vehicle, the location of another vehicle, or the size of another vehicle.

[0061] As an example, multiple frames may include information about at least one of the following: the frame acquisition period, surrounding traffic conditions, Global Positioning System (GPS) information, weather information (or type), and any combination thereof.

[0062] For example, processor 120 can classify a dataset into multiple bundles.

[0063] As an example, processor 120 can classify multiple frames contained in a dataset into different bundles. For instance, processor 120 can acquire multiple frames using sensor 130 and classify the dataset into multiple bundles based on the acquisition time of each frame.

[0064] As an example, processor 120 can classify a first frame acquired between a first time point and a second time point as a first bundle, and can classify a second frame acquired between a second time point and a third time point as a second bundle. The difference between the first and third time points can be a specific time (e.g., 30 minutes). In other words, processor 120 can classify a frame acquired after a specific time has elapsed from the acquisition time of the specific frame as a bundle different from the one acquired at that specific time.

[0065] As an example, processor 120 can adjust a specific time as a basis for bundle classification. For instance, if the split training data and the split evaluation data do not meet the first condition, processor 120 can determine whether the first condition is met based on multiple bundles that are reclassified by shortening the specific time.

[0066] For example, processor 120 can divide multiple bundles into training data or evaluation data.

[0067] As an example, processor 120 can identify multiple partitioning results that divide multiple bundles into training data and evaluation data. In other words, processor 120 can determine whether each of the multiple partitioning results that divide multiple bundles into training data and evaluation data satisfies a first condition. Multiple partitioning results can be defined, for example, as described below. Figure 4 The content in [the document / article].

[0068] For example, if the training data and the evaluation data meet the first condition, the processor 120 can perform an accuracy test on the training data and the evaluation data.

[0069] As an example, processor 120 can determine whether each of the above multiple partitioning results satisfies the first condition, and can perform an accuracy test on the results that satisfy the first condition and are partitioned into training data and evaluation data.

[0070] As an example, processor 120 can determine whether a first condition is met based on the number of specific objects (e.g., vehicles or cars) contained in the training data and evaluation data, respectively. For example, processor 120 can identify a first number of specific objects contained in a first frame of a first bundle classified as training data, and can identify a second number of specific objects contained in a second frame of a second bundle classified as evaluation data. Subsequently, if the ratio between the first and second numbers is within a specific error range relative to a predefined ratio, processor 120 can determine that the corresponding partitioning result (or the corresponding training data and the corresponding evaluation data) meets the first condition. For example, if the predefined ratio is 7:3, processor 120 can determine whether the ratio between the first and second numbers is within the specific error range relative to 7:3. The predefined ratio and / or the specific error range (e.g., 5%) can be settings that can be changed by the user and / or developer.

[0071] As an example, processor 120 can identify multiple partitioning results that divide multiple bundles into training data and evaluation data, and if none of the multiple partitioning results satisfy a first condition, the dataset can be classified into different bundles again. At this point, processor 120 can adjust (e.g., reduce) the benchmark used for partitioning into multiple bundles (e.g., the acquisition time difference between frames), thereby classifying the dataset into different bundles.

[0072] For example, if the results of the accuracy test meet the second condition, the processor 120 can train an artificial intelligence model based on the training data and the evaluation data, or it can evaluate the performance of the artificial intelligence model.

[0073] As an example, processor 120 can determine that a second condition is met if the difference between a first percentage of inclusion of each external object in the first frame of the first bundle divided as training data and a second percentage of inclusion of each external object in the second frame of the second bundle divided as evaluation data is less than or equal to a specific percentage. In other words, if the percentage of inclusion of the first object (e.g., a car) among the external objects included in the first frame is 70% relative to all objects, and the percentage of inclusion of the first object among the external objects included in the second frame is 65% relative to all objects, processor 120 can determine that the difference between these two percentages is less than or equal to a specific percentage (e.g., 5%) because the difference between these two percentages is 5%. In this way, processor 120 can determine that the second condition is met if the difference between the percentages of each other object (e.g., van, truck, bus, bicycle, and person) among the external objects included in the first and second frames is less than or equal to a specific percentage.

[0074] As an example, if the difference between the first standard deviation of the probability distribution of each external object (e.g., person (or pedestrian), car, truck, van, and bus) included in the first frame and the second standard deviation of the probability distribution of each external object included in the second frame is less than or equal to a specific value, then processor 120 can determine that a second condition is met. For example, processor 120 can calculate the first and second standard deviations of the probability distribution of each external object included in the first and second frames based on a bird's-eye view (BEV). For example, if the difference between the first and second standard deviations is less than or equal to a specific value, processor 120 can determine that the result of dividing the first and second frames is suitable for application to an artificial intelligence model. The specific value can be, for example, a setting that can be changed by the user and / or developer.

[0075] As an example, processor 120 can also determine whether the second condition is met based on benchmark information, which includes at least one of the following: weather information for both training and evaluation data; traffic congestion levels for both training and evaluation data; Global Positioning System (GPS) information for both training and evaluation data; the number of vehicles per frame for both training and evaluation data; and any combination thereof. For example, processor 120 can determine, based on the aforementioned benchmark information, whether some unnecessary duplicate data has been aggregated in the training or evaluation data, and if it is determined that some data has been aggregated, it can determine that the second condition is not met. For example, processor 120 can identify the traffic congestion level for each frame based on the number of vehicles per frame. For example, processor 120 can calculate the traffic congestion level based on whether the number of data labeled as cars in a particular frame is greater than a specific number (e.g., 40) in that particular frame.

[0076] For example, when determining whether the second condition described above is met, the processor 120 may determine the determination order based on a specific priority.

[0077] As an example, processor 120 may determine whether the second condition is met with a first priority based on a scheme using the percentage of external objects included, a second priority based on a scheme using the difference between the standard deviations of the probability distributions, and a third priority based on a scheme using reference information.

[0078] According to one example, sensor 130 can obtain information about at least one of the components contained in the vehicle, the driving environment of the vehicle, driving information of the vehicle, external information of the vehicle, and any combination thereof.

[0079] For example, sensor 130 can acquire sensor data. As an example, the sensor data may include sensor data acquired using LiDAR (Light Detection and Ranging). As an example, the sensor data may include information about external objects present around the vehicle.

[0080] Figure 2 An example is shown of the result of classifying multiple frames into multiple bundles in a vehicle control device according to an example of the present disclosure.

[0081] Referring to reference numerals 201 and 202 in the accompanying drawings, according to one example, a vehicle control device (e.g., Figure 1 The vehicle control unit 100 can use sensors to obtain a dataset containing multiple frames and can classify the dataset (or multiple frames) into at least one bundle.

[0082] Referring to reference numeral 201 in the accompanying drawings, according to one example, the vehicle control device can classify frames 1, 2, and 3 among a plurality of frames into bundle 1. The frames contained in bundle 1 can be defined as the first frame.

[0083] Referring to reference numeral 202, according to one example, the vehicle control device can classify frames 4, 5, 6, 7, 8, and 9 among a plurality of frames into bundle 2. Frames contained in bundle 2 can be defined as the second frame.

[0084] As an example, a vehicle control unit can classify a dataset into multiple bundles based on the acquisition time of each frame in multiple frames.

[0085] For example, if it is determined that a specific time (e.g., 30 minutes) has elapsed since the acquisition of frame 3, the vehicle control unit can classify frames acquired after frame 3 (e.g., frames 4 through 9) into a different bundle than frame 3. This specific time can be defined as a baseline time used to classify the dataset into multiple bundles.

[0086] For example, the vehicle control unit can adjust the base time. For example, if the above conditions are not met... Figure 1 The first condition described in the text allows the vehicle control unit to shorten the baseline time and reclassify the dataset into multiple bundles.

[0087] Figure 3 Examples of the results of dividing multiple bundles into training data or evaluation data in a vehicle control device according to the present disclosure are shown.

[0088] As an example, a vehicle control unit (e.g., Figure 1 The vehicle control device 100 in the middle can classify the dataset into multiple bundles, and can divide the classified multiple bundles into different data (e.g., training data or evaluation data).

[0089] For example, the vehicle control unit can divide bundle 1, bundle 3, bundle 5, bundle 6, bundle 7, bundle 8, and bundle 9 into first data (e.g., training data). The first data can be stored in a learning database 301.

[0090] For example, the vehicle control unit can divide bundles 2, 4, and 10 into second data (e.g., evaluation data). The second data can be stored in an evaluation database 302. The evaluation database 302 can be a memory (e.g., ...). Figure 1 Among the multiple databases contained in the memory 110, one is implemented as a separate database distinct from the learning database 301.

[0091] Figure 4 Examples of multiple division results of multiple bundles in a vehicle control device according to the present disclosure are shown.

[0092] As an example, a vehicle control unit (e.g., Figure 1 The vehicle control device 100 can identify multiple partitioning results that divide multiple bundles into training data 491 and evaluation data 492. Figure 4 The reference numerals 410, 420, and 430 shown in the figures can indicate a conceptual diagram based on multiple partitioning results. For example, a vehicle control unit can determine whether each of the multiple partitioning results satisfies a first condition. If none of the multiple partitioning results satisfy the first condition, the vehicle control unit can further classify the dataset into multiple bundles and determine whether the first condition is satisfied.

[0093] Hereinafter, for clarity of distinction, the reference numerals of the plurality of bundles corresponding to each of reference numerals 410, 420 and 430 are defined as different from each other, but bundles 1411, 1421 and 1431 shown in reference numerals 410, 420 and 430 may be the same bundles.

[0094] Referring to reference numeral 410, according to one example, the vehicle control unit may divide bundle 1 411 (or the frames contained in bundle 1 411) into training data 491, and may divide other bundles (e.g., bundle 2 412, bundle 3 413, ..., bundle 9 419) into evaluation data 492. For example, if the division result according to reference numeral 410 does not meet the first condition, the vehicle control unit may further examine the division result according to reference numeral 420.

[0095] Referring to reference numeral 420, according to one example, the vehicle control unit may divide bundle 1 421 and bundle 2 422 (or frames contained in bundle 1 421 and bundle 2 422) into training data 491, and may divide other bundles (e.g., bundle 3 423, ..., bundle 9 429) into evaluation data 492. For example, if the division result according to reference numeral 420 does not meet the first condition, the vehicle control unit may further examine the division result according to reference numeral 430.

[0096] Referring to reference numeral 430, according to one example, the vehicle control unit can divide bundle 1 431, bundle 2 432, and bundle 3 433 (or frames contained in bundle 1 431, bundle 2 432, and bundle 3 433) into training data 491, and can divide other bundles (e.g., bundle 4, ..., bundle 9 439) into evaluation data 492. For example, if the division result according to reference numeral 430 does not meet the first condition, the vehicle control unit can reclassify the dataset into multiple bundles based on different benchmarks (e.g., a benchmark with shortened benchmark time), and can determine whether the first condition is met based on the classification result.

[0097] Figure 5 An example is shown of the number of objects contained in the training and evaluation data according to the examples of this disclosure.

[0098] As an example, a vehicle control unit (e.g., Figure 1 The vehicle control unit 100 in the data can determine whether the second condition is met based on the percentage of external objects contained in each frame of different data.

[0099] For example, if the difference between a first percentage of each external object contained in a first frame of a first bundle of training data and a second percentage of each external object contained in a second frame of a second bundle of evaluation data is less than or equal to a certain percentage, the vehicle control unit can determine that the second condition is met.

[0100] Reference numeral 501 is a graph used to classify external objects in frames included in the training data, and reference numeral 502 is a graph used to classify external objects in frames included in the evaluation data.

[0101] For example, a vehicle control device can identify the percentage (or ratio) of the number of cars (203,977) in the training data out of the total number of external objects as the percentage of the first inclusion percentage corresponding to the cars.

[0102] For example, a vehicle control unit can identify the percentage (or ratio) of the number of cars (34,424) in the evaluation data out of the total number of external objects as the percentage of inclusion corresponding to cars in a second inclusion percentage.

[0103] For example, if the difference between the percentage of cars included in the training data and the evaluation data is less than or equal to a specific percentage, the vehicle control unit can similarly determine whether the difference between the percentages of each other external object (e.g., vans, trucks, buses, cyclists, and pedestrians) included is less than or equal to a specific percentage.

[0104] Through the above determination operation, if the difference between the first inclusion percentage and the second inclusion percentage of all external objects is determined to be less than or equal to a specific percentage, the vehicle control device can determine that the second condition is met.

[0105] Figures 6A to 6D It is an illustration showing a specific object contained in the training and evaluation data in a specific technique (e.g., bird's-eye view (BEV)) according to an example of this disclosure.

[0106] Reference Figure 6A According to one example, vehicle control devices (e.g., Figure 1 The vehicle control unit 100 in the training data can calculate the standard deviation of the probability distribution of the first external object in the frames contained in the training data and the evaluation data, based on a bird's-eye view (BEV) corresponding to the first external object (e.g., a bus).

[0107] For example, reference numeral 611 may be the BEV of the first external object in the training data, and reference numeral 612 may be the BEV of the first external object in the evaluation data.

[0108] Reference Figure 6B According to one example, the vehicle control unit can calculate the standard deviation of the probability distribution of the second external object in the frames contained in the training data and the evaluation data, based on the BEV corresponding to the second external object (e.g., a car).

[0109] For example, reference numeral 621 may be the BEV of the second external object in the training data, and reference numeral 622 may be the BEV of the second external object in the evaluation data.

[0110] Reference Figure 6C According to one example, the vehicle control unit can calculate the standard deviation of the probability distribution of the third external object in the frames contained in the training data and the evaluation data, based on the BEV corresponding to the third external object (e.g., a pedestrian).

[0111] For example, reference numeral 631 may be the BEV of a third external object in the training data, and reference numeral 632 may be the BEV of a third external object in the evaluation data.

[0112] Reference Figure 6D According to one example, the vehicle control unit can calculate the standard deviation of the probability distribution of the fourth external object in the frames contained in the training data and the evaluation data, based on the BEV corresponding to the fourth external object (e.g., a cyclist).

[0113] For example, reference numeral 641 may be the BEV of the fourth external object in the training data, and reference numeral 642 may be the BEV of the fourth external object in the evaluation data.

[0114] Figure 7 An example of a vehicle control method according to this disclosure is shown. For convenience, Figure 7 This can be described by an example of steps being performed by a processor (e.g., control circuitry). Figure 7 One, some, or all of the steps, or a portion thereof, may be performed by one or more other circuits. Figure 7 One or more steps may be omitted, performed in a different order, and / or modified in other ways, and / or one or more additional steps may be added.

[0115] According to the example, vehicle control devices (e.g., Figure 1 The vehicle control device 100 in the middle can perform Figure 7 The operation disclosed herein. For example, at least some components included in a vehicle control device (e.g., Figure 1 The memory 110, processor 120, and sensor 130 in the memory can be configured to perform Figure 7 The operations within.

[0116] In the following examples, operations S710 to S760 can be executed sequentially, but not necessarily in that order. For example, the order of the operations can be changed, and at least two operations can be executed in parallel. Furthermore, in conjunction with the above... Figure 7 Content that corresponds to or repeats the described content can be briefly described or omitted.

[0117] According to one example, in the S710, the vehicle control unit can obtain a dataset.

[0118] For example, a vehicle control unit can use sensors to acquire a dataset containing multiple frames for autonomous driving control of the vehicle. As an example, this dataset could include sensor data acquired using LiDAR.

[0119] As an example, in the S720, the vehicle control unit can classify a dataset into multiple bundles.

[0120] For example, a vehicle control unit can classify multiple frames contained in a dataset into different bundles.

[0121] According to one example, in the S730, the vehicle control unit can divide multiple bundles into training data or evaluation data.

[0122] For example, a vehicle control unit can divide multiple bundles into two groups and identify these two groups as training data and evaluation data, respectively.

[0123] According to one example, in S740, the vehicle control unit can determine whether the training data and evaluation data meet the first condition.

[0124] For example, if the ratio between the number of a specific object (e.g., a car) in the training data and the number of a specific object (e.g., a car) in the evaluation data is within a certain error range relative to a predefined ratio, the vehicle control unit can determine that the first condition is met.

[0125] For example, if the training data and evaluation data meet the first condition (e.g., S740 - yes), the vehicle control unit can execute S750.

[0126] For example, if the training data and evaluation data do not meet the first condition (e.g., S740 - No), the vehicle control unit may repeat S720.

[0127] According to one example, in S750, the vehicle control unit can determine whether the accuracy test results of the training data and evaluation data meet the second condition.

[0128] For example, the vehicle control unit may determine whether the second condition is met based on a scheme that uses percentages of external objects containing training data and evaluation data respectively, a scheme that uses the difference between the standard deviations of the probability distributions, and / or a scheme that uses benchmark information.

[0129] For example, if the accuracy test results of the training data and the evaluation data meet the second condition (e.g., S750 - Yes), the vehicle control unit can execute S760.

[0130] For example, if the accuracy test results of the training data and the evaluation data do not meet the second condition (e.g., S750 - No), the vehicle control unit may repeat S720.

[0131] As an example, in the S760, the vehicle control unit can train an artificial intelligence model or evaluate the performance of an artificial intelligence model based on training and evaluation data.

[0132] For example, vehicle control devices can train artificial intelligence models, thereby improving the object recognition performance of these models.

[0133] Figure 8 An example of a computational system for a vehicle control method is shown according to an example of this disclosure.

[0134] Reference Figure 8 The computing system 1000 for the vehicle control method may include at least one processor 1100, a memory 1300, a user interface input device 1400, a user interface output device 1500, a storage device 1600, and a network interface 1700, which are interconnected via a bus 1200.

[0135] Processor 1100 may be a central processing unit (CPU) or a semiconductor device that processes instructions stored in memory 1300 and / or storage device 1600. Memory 1300 and storage device 1600 may include various types of volatile or non-volatile storage media. For example, memory 1300 may include ROM (Read-Only Memory) 1310 and RAM (Random Access Memory) 1320.

[0136] Therefore, the operation of the methods or algorithms described in conjunction with the examples disclosed in this specification can be directly implemented using hardware modules, software modules, or a combination of hardware and software modules executed by processor 1100. Software modules can reside in storage media (i.e., memory 1300 and / or storage device 1600), such as RAM, flash memory, ROM, EPROM, EEPROM, registers, hard disks, removable disks, and CD-ROMs.

[0137] An exemplary storage medium may be coupled to processor 1100. Processor 1100 may read information from the storage medium and may write information to the storage medium. Alternatively, the storage medium may be integrated with processor 1100. The processor and storage medium may reside in an application-specific integrated circuit (ASIC). The ASIC may reside within the user terminal. In another case, the processor and storage medium may reside as separate components in the user terminal.

[0138] This disclosure is made to address the aforementioned problems in the prior art, while maintaining the advantages achieved by the prior art.

[0139] One example of this disclosure provides a vehicle control device for dividing a dataset into a specific proportion of training and evaluation data based on a user- or developer-defined benchmark, in order to improve the performance of an artificial intelligence model (or deep learning network).

[0140] Another example of this disclosure provides a vehicle control device that is implemented to efficiently and quickly partition a dataset to minimize unnecessary load and reduce resources and time.

[0141] The technical problems to be solved by this disclosure are not limited to those described above, and any other technical problems not mentioned herein will be clearly understood by those skilled in the art from the following description.

[0142] According to one example of this disclosure, a vehicle control device may include a memory storing at least one instruction and a processor operatively connected to the memory. For example, when executed by the processor, the at least one instruction may cause the vehicle control device to: obtain a dataset containing multiple frames for driving control of the vehicle; classify the dataset into multiple bundles; divide the multiple bundles into training data or evaluation data; perform an accuracy test on the training data and evaluation data if the training data and evaluation data satisfy a first condition; and train an artificial intelligence model or evaluate the performance of the artificial intelligence model based on the training data and evaluation data if the result of performing the accuracy test satisfies a second condition.

[0143] According to one example, the vehicle control unit may also include sensors. For example, at least one instruction, when executed by a processor, may cause the vehicle control unit to: input sensor data obtained using the sensors into an artificial intelligence model to detect objects present outside the vehicle during driving control.

[0144] According to one example, multiple frames may include at least one of the following: the category of the external object, the location of the external object, the size of the external object, the acquisition period, the surrounding traffic environment, global positioning system (GPS) information, weather information, and any combination thereof.

[0145] According to one example, at least one instruction, when executed by a processor, can cause a vehicle control device to: identify a first number of specific objects contained in a first frame of a first bundle of training data; identify a second number of specific objects contained in a second frame of a second bundle of evaluation data; and determine that a first condition is satisfied if the ratio between the first number and the second number is within a certain error range relative to a predefined ratio.

[0146] According to one example, at least one instruction, when executed by a processor, can enable a vehicle control unit to: acquire multiple frames using sensors; and classify the dataset into multiple bundles based on the acquisition time of each of the multiple frames.

[0147] According to one example, at least one instruction, when executed by the processor, can cause the vehicle control unit to: shorten the baseline time for classifying the dataset into multiple bundles if the first condition is not met, and classify the dataset into multiple bundles again.

[0148] According to one example, at least one instruction, when executed by a processor, can cause the vehicle control unit to: identify multiple partitioning results that divide multiple bundles into training data and evaluation data; and if none of the multiple partitioning results satisfy a first condition, then reclassify the dataset into different multiple bundles.

[0149] According to one example, at least one instruction, when executed by a processor, can cause the vehicle control unit to determine that a second condition is satisfied if the difference between a first percentage of each external object contained in a first frame of a first bundle divided into training data and a second percentage of each external object contained in a second frame of a second bundle divided into evaluation data is less than or equal to a certain percentage.

[0150] According to one example, at least one instruction, when executed by the processor, can cause the vehicle control unit to determine that a second condition is satisfied if the difference between the first standard deviation of the probability distribution of each external object contained in the first frame and the second standard deviation of the probability distribution of each external object contained in the second frame is less than or equal to a specific value.

[0151] According to one example, at least one instruction, when executed by the processor, can cause the vehicle control device to determine whether the second condition is satisfied based on at least one of the following: weather information of the training data and evaluation data respectively, traffic congestion of the training data and evaluation data respectively, global positioning system (GPS) information of the training data and evaluation data respectively, the number of vehicles per frame of the training data and evaluation data respectively, and any combination thereof.

[0152] According to another example of this disclosure, a vehicle control method may include: obtaining a dataset containing multiple frames for driving control of the vehicle by a processor; classifying the dataset into multiple bundles by the processor; dividing the multiple bundles into training data or evaluation data by the processor; performing an accuracy test on the training data and evaluation data by the processor if the training data and evaluation data satisfy a first condition; and training an artificial intelligence model or evaluating the performance of the artificial intelligence model by the processor based on the training data and evaluation data if the result of performing the accuracy test satisfies a second condition.

[0153] According to one example, the vehicle control method may also include: the processor inputting sensor data obtained using sensors into an artificial intelligence model to detect objects present outside the vehicle during driving control.

[0154] According to one example, multiple frames may include at least one of the following: the category of the external object, the location of the external object, the size of the external object, the acquisition period, the surrounding traffic environment, global positioning system (GPS) information, weather information, and any combination thereof.

[0155] According to one example, the vehicle control method may further include: identifying by the processor a first number of specific objects contained in a first frame of a first bundle of training data; identifying by the processor a second number of specific objects contained in a second frame of a second bundle of evaluation data; and determining by the processor that a first condition is satisfied if the ratio between the first number and the second number is within a certain error range relative to a predefined ratio.

[0156] According to one example, the vehicle control method may also include: acquiring multiple frames by a processor using sensors; and classifying the dataset into multiple bundles by the processor based on the acquisition time of each of the multiple frames.

[0157] According to one example, the vehicle control method may further include: if the first condition is not met, the processor shortens the baseline time for classifying the dataset into multiple bundles, and the processor classifies the dataset into multiple bundles again.

[0158] According to one example, the vehicle control method may further include: having a processor identify multiple partitioning results that divide multiple bundles into training data and evaluation data; and having the processor reclassify the dataset into different multiple bundles if none of the multiple partitioning results satisfy a first condition.

[0159] According to one example, the vehicle control method may further include: if the difference between a first percentage of each external object contained in a first frame in a first bundle of training data and a second percentage of each external object contained in a second frame in a second bundle of evaluation data is less than or equal to a specific percentage, then the processor determines that a second condition is satisfied.

[0160] According to one example, the vehicle control method may further include: if the difference between a first standard deviation of the probability distribution of each external object contained in the first frame and a second standard deviation of the probability distribution of each external object contained in the second frame is less than or equal to a specific value, then the processor determines that a second condition is satisfied.

[0161] According to one example, the vehicle control method may further include: the processor determining whether the second condition is met based on at least one of the following: weather information of the training data and evaluation data respectively, traffic congestion of the training data and evaluation data respectively, global positioning system (GPS) information of the training data and evaluation data respectively, the number of vehicles per frame of the training data and evaluation data respectively, and any combination thereof.

[0162] The effects of a vehicle control device and method according to an example of this disclosure will now be described.

[0163] Examples of this disclosure may provide a vehicle control device for dividing a dataset into a specific proportion of training and evaluation data based on a benchmark defined by a user or developer, in order to improve the performance of an artificial intelligence model (or deep learning network).

[0164] Examples of this disclosure may provide a vehicle control device that is implemented to efficiently and quickly partition a dataset to minimize unnecessary load and reduce resources and time.

[0165] In addition, various effects that can be directly or indirectly determined through this disclosure may be provided.

[0166] Although the present disclosure has been described above with reference to examples and drawings, the present disclosure is not limited thereto. Various modifications and alterations can be made to the present disclosure by those skilled in the art without departing from the spirit and scope of the present disclosure as claimed in the following claims.

[0167] Therefore, the examples in this disclosure are not intended to limit the technical spirit of this disclosure, but are provided for illustrative purposes only. The scope of this disclosure should be interpreted in accordance with the appended claims, and all technical ideas within the scope of the claims should be included within the scope of this disclosure.

Claims

1. A device for controlling the movement of a vehicle, the device comprising: A memory that stores at least one instruction; and A processor operatively coupled to the memory, Wherein, when executed by the processor, the at least one instruction is configured to cause the device to: Obtain a dataset containing multiple frames used for vehicle driving control; The dataset is classified into multiple bundles; The multiple bundles are divided into training data and evaluation data; Based on the fact that the training data and the evaluation data meet the first condition, an accuracy test is performed on the training data and the evaluation data; Based on the accuracy test meeting the second condition, train the artificial intelligence model or evaluate the performance of the artificial intelligence model; Based on the trained or evaluated performance of the artificial intelligence model, an output signal is generated; and The vehicle's movement is controlled based on the signal.

2. The apparatus according to claim 1, further comprising: sensor, Wherein, when executed by the processor, the at least one instruction is configured to cause the device to: Sensor data obtained using the sensors is input into the artificial intelligence model to detect objects present outside the vehicle during driving control.

3. The apparatus according to claim 1, wherein, The plurality of frames includes at least one of the following: The category of external objects; The location of the external object; The size of the external object; Acquisition period, wherein the acquisition period corresponds to the duration for which frame data is collected within the bundle; Surrounding traffic information; Global Positioning System (GPS) information; and Weather information.

4. The apparatus according to claim 1, wherein, When executed by the processor, the at least one instruction is configured to cause the device to: Identify a first number of specific objects contained in a first frame, wherein the first frame is divided into a first bundle of the training data in the plurality of bundles; Identify a second number of the specific objects contained in the second frame, wherein the second frame is divided into a second bundle of the evaluation data within the plurality of bundles; and Based on the fact that the ratio between the first quantity and the second quantity is within a certain error range relative to a predefined ratio, it is determined that the first condition is met.

5. The apparatus according to claim 2, wherein, When executed by the processor, the at least one instruction is configured to cause the device to: The multiple frames are obtained using the sensor; as well as Based on the acquisition time of each of the multiple frames, the dataset is classified into the multiple bundles.

6. The apparatus according to claim 5, wherein, When executed by the processor, the at least one instruction is configured to cause the device to: Based on the fact that the first condition is not met, the baseline time for classifying the dataset into the plurality of bundles is shortened, and the dataset is classified into the plurality of bundles again.

7. The apparatus according to claim 1, wherein, When executed by the processor, the at least one instruction is configured to cause the device to: Identify multiple partitioning results that divide the multiple bundles into the training data and the evaluation data; as well as Since none of the multiple partitioning results satisfy the first condition, the dataset is further classified into multiple different bundles.

8. The apparatus according to claim 1, wherein, When executed by the processor, the at least one instruction is configured to cause the device to: Based on the fact that the difference between the first inclusion percentage and the second inclusion percentage is less than or equal to a specific percentage, it is determined that the second condition is satisfied, wherein: The first percentage corresponds to the proportion of the first frame containing each external object, the first frame being divided into the first bundle of the training data in the plurality of bundles, and The second percentage corresponds to the proportion of the second frame containing each external object, which is divided into the second bundle of the evaluation data in the plurality of bundles.

9. The apparatus according to claim 8, wherein, When executed by the processor, the at least one instruction is configured to cause the device to: Based on the fact that the difference between the first standard deviation and the second standard deviation is less than or equal to a specific value, it is determined that the second condition is satisfied, wherein: The first standard deviation corresponds to the standard deviation of the probability distribution of each external object contained in the first frame, and The second standard deviation corresponds to the standard deviation of the probability distribution of each external object contained in the second frame.

10. The apparatus according to claim 8, wherein, When executed by the processor, the at least one instruction is configured to cause the device to: The determination of whether the second condition is met is also based on at least one of the following: The weather information for the training data and the evaluation data, respectively; The traffic congestion levels of the training data and the evaluation data, respectively; The Global Positioning System (GPS) information of the training data and the evaluation data, respectively; as well as The number of vehicles per frame in both the training data and the evaluation data.

11. A method performed by a device for controlling the movement of a vehicle, the method comprising the steps of: Obtain a dataset containing multiple frames used for vehicle driving control; The dataset is classified into multiple bundles; The multiple bundles are divided into training data and evaluation data; Based on the fact that the training data and the evaluation data meet the first condition, an accuracy test is performed on the training data and the evaluation data; Based on the accuracy test meeting the second condition, train the artificial intelligence model or evaluate the performance of the artificial intelligence model; Based on the performance of the trained or evaluated artificial intelligence model, an output signal is generated; as well as Based on the signal, the vehicle's movement is controlled.

12. The method of claim 11, further comprising the step of: Sensor data obtained using the vehicle's sensors is input into the artificial intelligence model to detect objects present outside the vehicle during driving control.

13. The method according to claim 11, wherein, The plurality of frames includes at least one of the following: The category of external objects; The location of the external object; The size of the external object; Acquisition period, wherein the acquisition period corresponds to the duration for which frame data is collected within the bundle; Surrounding traffic information; Global Positioning System (GPS) information; and Weather information.

14. The method of claim 11, further comprising the step of: Identify a first number of specific objects contained in a first frame, wherein the first frame is divided into a first bundle of the training data in the plurality of bundles; Identify a second number of the specific objects contained in the second frame, wherein the second frame is divided into a second bundle of the evaluation data within the plurality of bundles; and Based on the fact that the ratio between the first quantity and the second quantity is within a certain error range relative to a predefined ratio, it is determined that the first condition is met.

15. The method of claim 12, further comprising the step of: The multiple frames are obtained using the sensor; as well as Based on the acquisition time of each of the multiple frames, the dataset is classified into the multiple bundles.

16. The method of claim 15, further comprising the step of: Based on the fact that the first condition is not met, the baseline time for classifying the dataset into the plurality of bundles is shortened, and the dataset is classified into the plurality of bundles again.

17. The method of claim 11, further comprising the step of: Identify multiple partitioning results that divide the multiple bundles into the training data and the evaluation data; as well as Since none of the multiple partitioning results satisfy the first condition, the dataset is further classified into multiple different bundles.

18. The method of claim 11, further comprising the step of: Based on the fact that the difference between the first inclusion percentage and the second inclusion percentage is less than or equal to a specific percentage, it is determined that the second condition is satisfied, wherein: The first percentage corresponds to the proportion of the first frame containing each external object, the first frame being divided into a first bundle of the training data within a plurality of bundles; and The second percentage corresponds to the proportion of the second frame containing each external object, which is divided into the second bundle of the evaluation data in a plurality of bundles.

19. The method of claim 18, further comprising the step of: Based on the fact that the difference between the first standard deviation and the second standard deviation is less than or equal to a specific value, it is determined that the second condition is satisfied, wherein: The first standard deviation corresponds to the standard deviation of the probability distribution of each external object contained in the first frame; and The second standard deviation corresponds to the standard deviation of the probability distribution of each external object contained in the second frame.

20. The method of claim 18, further comprising the step of: The determination of whether the second condition is met is also based on at least one of the following: The weather information for the training data and the evaluation data, respectively; The traffic congestion levels of the training data and the evaluation data, respectively; The Global Positioning System (GPS) information of the training data and the evaluation data, respectively; as well as The number of vehicles per frame in both the training data and the evaluation data.

Images