KR102600515B1 - Method, apparatus, computer-readable storage medium and computer program for generating driving control data - Google Patents

Abstract

According to an embodiment, a device that communicates with an edge node and generates driving control data includes: a data acquisition unit that acquires driving data around the device; and a driving control data generator that generates driving control data using the driving data or using the driving data and the edge data, according to a control signal generated based on edge data obtained from the edge node. .

Description

Driving control data generation method, device, computer-readable recording medium and computer program {METHOD, APPARATUS, COMPUTER-READABLE STORAGE MEDIUM AND COMPUTER PROGRAM FOR GENERATING DRIVING CONTROL DATA}

The present invention relates to a method and device for generating driving control data, a computer-readable recording medium, and a computer program.

Machine learning-based autonomous driving technology performs control by processing data obtained from sensors in real time.

Specifically, in an edge computing environment, a device (eg, autonomous driving device) may generate driving control data by receiving edge data (or computational resource data) obtained from a sensor included in an edge node.

However, in an edge computing environment, data synchronization between edge nodes and devices is limited. Therefore, due to unstable communication between the edge and the device or limitations in computational resources, including the network, edge data (or computational resource data) obtained from sensors included in the edge node cannot be delivered to the device (or control subject) in real time. A case may occur, in which case the device (or control subject) had to make a control decision based on previous data that was not updated.

However, when the device (control subject) does not receive edge data (or computational resource data) from the edge node and makes control decisions based on previous data that has not been updated, there is a problem that the control performance of the device is lowered. .

Korea Patent Publication, No. 10-2016-0049962 (published May 10, 2016)

The problem to be solved by the present invention is to provide a method and device for generating driving control data, a computer-readable recording medium, and a computer program.

In addition, through these driving control data generation methods, devices, computer-readable recording media, and computer programs, control signals are generated through an orchestrator previously learned through reinforcement learning even in situations where the communication environment between the edge node and the device is unstable. Accordingly, the ability to generate driving control data may be included in the problem to be solved by the present invention.

However, the problems to be solved by the present invention are not limited to those mentioned above, and other problems to be solved that are not mentioned can be clearly understood by those skilled in the art to which the present invention pertains from the description below. will be.

A device according to an embodiment of the present invention is a device that communicates with an edge node and generates driving control data, comprising: a data acquisition unit that acquires driving data around the device; and a driving control data generator that generates driving control data using the driving data or using the driving data and the edge data, according to a control signal generated based on edge data obtained from the edge node. .

Additionally, the control signal may be generated by the edge node.

Additionally, a first auto-encoder that generates first driving control data using the driving data; and a second auto-encoder that generates second driving control data using the driving data and the edge data, wherein the driving control data generator generates the first auto-encoder and the second driving control data according to the control signal. The driving control data can be generated using any one of the auto encoders.

In addition, when the control signal indicates that the driving control data is generated using the first auto encoder, the driving control data generator generates the first driving control data using the first auto encoder, and , when the control signal indicates that the driving control data is generated using the second auto encoder, the second driving control data may be generated using the second auto encoder.

In addition, when the first autoencoder receives driving data for learning and first correct driving control data as label data for the driving data for learning, it encodes the driving data for learning and outputs the first correct driving control data. It is already trained to do so, and when the second auto encoder receives learning edge data, learning driving data, and second correct driving control data as label data for the learning edge data and learning driving data, the learning edge data It may be previously learned to output the second correct driving control data by encoding data and the learning driving data.

In addition, it may further include an orchestrator that receives the edge data from the edge node and generates the control signal.

In addition, the orchestrator is pre-trained to output a learning control signal for selecting one of the first auto encoder and the second auto encoder when receiving edge data for learning, and the learning control signal and the correct answer control signal By using a loss function for the difference, when the learning edge data is input, it can be further learned to output the learning control signal.

Additionally, the data acquisition unit may acquire the driving data using at least one of a camera and a lidar sensor.

A method of generating driving control data according to an embodiment of the present invention is performed by a device that communicates with an edge node and generates driving control data, comprising: acquiring driving data around the device; and generating driving control data using the driving data or using the driving data and the edge data according to a control signal generated based on edge data obtained from the edge node.

Additionally, the control signal may be generated by the edge node.

In addition, the step of generating the driving control data includes a first autoencoder that generates first driving control data using the driving data according to the control signal, and a second driving control data using the driving data and the edge data. The driving control data can be generated using any one of the second auto encoders that generate control data.

Additionally, when the control signal indicates that the driving control data is generated using the first auto encoder, the first driving control data is generated using the first auto encoder, and the control signal is used to generate the driving control data using the first auto encoder. 2 When indicating that the driving control data is generated using an auto encoder, the second driving control data can be generated using the second auto encoder.

In addition, when the first autoencoder receives driving data for learning and first correct driving control data as label data for the driving data for learning, it encodes the driving data for learning and outputs the first correct driving control data. It is already trained to do so, and when the second auto encoder receives learning edge data, learning driving data, and second correct driving control data as label data for the learning edge data and learning driving data, the learning edge data It may be previously learned to output the second correct driving control data by encoding data and the learning driving data.

Additionally, the method may further include receiving edge data from the edge node and generating the control signal.

In addition, the step of generating the control signal is pre-trained so that when learning edge data is input, a learning control signal for selecting one of the first auto encoder and the second auto encoder is output, and the learning control signal and By using a loss function for the difference from the correct answer control signal, when the learning edge data is input, the control signal can be generated using an orchestrator that is further trained to output the learning control signal.

Additionally, in the step of acquiring the driving data, the driving data may be acquired using at least one of a camera and a LiDAR sensor.

According to the device according to an embodiment of the present invention, a suitable control signal is generated even in an environment where communication between an edge node and the device is not smooth or limited, and the driving data obtained from the device is used according to the generated control signal, or the driving data and Driving control data can be generated using edge data obtained from edge nodes.

Additionally, devices according to embodiments of the present invention can be used in robots, autonomous drones, and smart factories that operate based on edge computing.

However, the effects that can be obtained from the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below. You will be able to.

Figure 1 is a diagram to explain the communication environment between an edge node and a device.
Figure 2 is a configuration diagram of a device according to an embodiment of the present invention.
Figure 3 is a configuration diagram showing a device according to another embodiment of the present invention.
Figure 4 is a diagram for explaining an orchestrator according to an embodiment of the present invention.
Figure 5 is a diagram for explaining reinforcement learning of an orchestrator according to an embodiment of the present invention.
FIGS. 6A and 6B are diagrams to explain how driving control data is generated by a device according to an embodiment of the present invention.
7 is an exemplary flowchart of the procedure of a method for generating driving control data according to an embodiment of the present invention.

The advantages and features of the present invention and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms. The present embodiments are merely provided to ensure that the disclosure of the present invention is complete and to be understood by those skilled in the art. It is provided to fully inform those who have the scope of the invention, and the present invention is only defined by the scope of the claims.

In describing embodiments of the present invention, if a detailed description of a known function or configuration is judged to unnecessarily obscure the gist of the present invention, the detailed description will be omitted. The terms described below are terms defined in consideration of functions in the embodiments of the present invention, and may vary depending on the intention or custom of the user or operator. Therefore, the definition should be made based on the contents throughout this specification.

Figure 1 is a diagram to explain the communication environment between an edge node and a device.

Referring to FIG. 1, the edge node 10 includes a sensor (not shown), acquires edge data through the sensor, and transmits the obtained edge data to the device (or device) 100.

For example, the edge node 10 may include a radar, acquire edge data using the radar, and transmit the edge data obtained using the radar to the device 100.

The apparatus (or device) 100 includes a sensor (not shown), and can obtain driving data around the apparatus 100 (or within a predetermined distance from the apparatus) through the sensor. For example, the device 100 may include a camera or a LiDAR sensor, and may obtain driving data within a predetermined distance from the device using the camera or the LiDAR. For example, the device 100 is an autonomous driving device and may include, for example, a self-driving drone or a self-driving car.

Here, the driving data may be at least one of LiDAR data and image data regarding obstacles within a predetermined distance from the device 100.

The device 100 may receive edge data obtained from the edge node 10. Here, the edge data that the device 100 receives from the edge node 10 may be computational resource data used to generate driving control data of the device, and the device 100 receives the edge data from the edge node 10. Driving control data can be generated using this. Here, the driving control data may be control data for autonomous driving of the device 100.

However, when the communication between the edge node 10 and the device 100 is unstable (or the communication network is unstable), the device 100 does not receive the edge data obtained from the edge node 10, and accordingly, the driving control Problems may arise when generating data.

Accordingly, the device 100 according to an embodiment of the present invention generates a suitable control signal even in an environment where communication between the edge node 10 and the device 100 is not smooth or is limited, and the device ( Driving control data can be generated using the driving data obtained from 100) or using the driving data and edge data obtained from the edge node 10. Hereinafter, the device 100 according to an embodiment of the present invention will be described in detail with reference to FIG. 2.

Figure 2 is a configuration diagram of a device according to an embodiment of the present invention.

Referring to FIG. 2, the device 100 according to an embodiment includes a data acquisition unit 110, a first auto encoder 120, a second auto encoder 130, a driving control data generator 140, and an orchestrator. It may include (150).

The data acquisition unit 110 may acquire driving data around the device 100. For example, the data acquisition unit 110 may acquire driving data using at least one of a camera and a LiDAR sensor.

Here, the driving data may be at least one of LiDAR data and image data regarding obstacles within a predetermined distance from the device 100.

The first auto encoder 120 may generate first driving control data using driving data. In more detail, the first auto encoder 120 is training driving data and label data for the learning driving data, and when receiving the first correct answer driving control data, encodes the learning driving data to output the first correct answer driving control data. It may have already been learned.

The second auto encoder 130 may generate second driving data using driving data and edge data. In more detail, the second auto encoder 130 encodes the learning edge data and the learning driving data as label data for the learning edge data and the learning driving data, and generates the second correct answer driving control data. It may have already been learned to print.

The driving control data generator 140 may generate driving control data using driving data or using driving data and edge data according to a control signal generated based on edge data obtained from an edge node.

In more detail, the driving control data generator 140 may generate driving control data using either the first auto encoder 120 or the second auto encoder 130 according to the generated control signal.

For example, when the generated control signal indicates that the driving control data is generated using the first auto encoder 120, the driving control data generator 140 uses the first auto encoder 120 to generate the driving control data. 1 When driving control data is generated and the generated control signal indicates that driving control data is generated using the second auto encoder 130, the second driving control data is generated using the second auto encoder 130. can be created.

The orchestrator 150 may receive edge data from the edge node 10 and generate a control signal.

As an embodiment, the orchestrator 150 is pre-trained to output a learning control signal for selecting one of the first auto encoder and the second auto encoder when receiving an edge for learning, and outputs a learning control signal and a correct answer control signal. By using a loss function for the difference, when edge data for learning is input, it may be further learned so that a control signal for learning is output.

Figure 3 is a configuration diagram showing a device 100 according to another embodiment of the present invention.

Referring to FIG. 3, the device 100 according to another embodiment of the present invention may not include the orchestrator 150. In this case, the device 100 according to another embodiment of the present invention may be used in a communication unit (not shown). A control signal generated by the orchestrator 30 included in the edge node 10 can be received.

Hereinafter, the orchestrator will be described in more detail with reference to FIGS. 4 and 5. Here, the orchestrator described in the description with reference to FIGS. 4 and 5 may be either the orchestrator 150 included in the device 100 or the orchestrator 30 included in the edge node 10.

Figure 4 is a diagram for explaining an orchestrator according to an embodiment of the present invention.

Referring to FIG. 4, the orchestrator may generate a control signal based on edge data obtained from the edge node 10. Here, the control signal refers to a method of generating driving control data in the device 100.

The orchestrator may be pre-trained based on reinforcement learning to select the most optimized control signal among a plurality of control signals to determine whether to utilize the edge data received from the edge node 10.

In more detail, a method of reinforcement learning to generate a control signal in an orchestrator according to an embodiment will be described.

First, (1) the orchestrator is a low-level control signal (control policy) that performs MDP (Markov Decision Process) creates .

(2) Lower control signal Through, samples of state s follow the optimal state distribution Collect.

(3) Sample Raw autoencoder (first autoencoder or second autoencoder) By learning, the map of latent variables z and s obtain.

(4) Projection map corresponding to each update cycle t through time state A random observation corresponding to and create a neural network cast Supervised learning is performed on the autoencoder (first autoencoder or second autoencoder) to respond to .

(5) Control signal to observe latent variable z and take action a creates .

(6) Finally, control signal can be obtained.

A method of reinforcement learning to generate a control signal in an orchestrator according to another embodiment will be described.

First, (1) the orchestrator learns the control signal generated using edge data and driving data acquired from the device 100.

(2) Afterwards, samples of data (edge data and driving data) observed through the learned control signal ( ) to obtain.

(3) observed samples ( ) to learn the auto encoder (first auto encoder or second auto encoder).

(4) Data from which part (or all) of the edge data was removed using the auto encoder learned in step (3) ( ), and the encoding result of the observation (output value encoded by inputting the observed sample into the autoencoder, ) is used as a label to perform supervised learning on the autoencoder.

(5) (4) As a supervised learning result, an autoencoder learning model ( ) to obtain.

(6) Learning model in an environment with no delay in receiving edge data The result of encoding using Observe Result Policy Learn with reinforcement learning. in other words, is the encoding result Take input and act can be inferred.( )

Finally, (7) the supervised learning results obtained in step (5) and, It can be learned by learning to obtain a control policy. At this time, the encoding result of the auto encoder is You can infer behavior by passing it to . In addition, encoding is performed using only driving data to take action. can be inferred.

Figure 5 is a diagram for explaining reinforcement learning of an orchestrator according to an embodiment of the present invention.

Referring to FIG. 5, the orchestrator can select a control policy only when connected to an edge node, which reduces sample efficiency, so learning can be performed in a model environment and a simulator environment, respectively.

At this time, the learning of the orchestrator must be performed in an environment where there is no delay in receiving edge data from the edge node 10. This is because if learning is performed with a delay in receiving edge data, This is because learning is not performed properly because it does not satisfy the mathematical model of the reinforcement learning environment, MDP (Markov Decision Process) or POMDP (Partial Observability MDP).

In addition, when learning of the orchestrator is performed in an environment where there is no delay in receiving edge data from the edge node 10, various reinforcement learning algorithms such as Soft Actor Critic (SAC) and Proximal Policy Optimization (PPO) can be effectively used. You can use it.

On the other hand, since the orchestrator's time is limited to updates, there is a problem that the orchestrator's learning is delayed by the average delay time. To this end, model-based learning is introduced for data efficiency. Since general model-based learning has a model bias problem, aggregated policy technology borrowed from federated learning techniques is introduced. This aggregate policy can be obtained as a weighted average of the policy parameters obtained from model-based learning and the parameters obtained by attaching model free learning.

More specifically, the orchestrator performs training on the simulator (parameters ), after performing learning in the model environment (parameters ), weights preset by the federated learning technique ( ), the parameters in each learning process ( , ) can be used by weighted averaging as shown in Equation 1 below.

Meanwhile, the orchestrator sets parameters for each simulator learning and learning process in the model environment ( , ) can be updated as shown in Equation 2 below.

At this time, the control policy selected by the orchestrator may be maintained until a later control policy is selected.

FIGS. 6A and 6B are diagrams to explain how driving control data is generated by a device according to an embodiment of the present invention. In more detail, FIG. 6A is a diagram showing the edge node 10 including the orchestrator 30, and FIG. 6B is a diagram showing the device 100 including the orchestrator 150 according to an embodiment of the present invention.

Referring to FIGS. 6A and 6B , first (1) the edge node 10 may acquire edge data using a sensor (eg, radar).

Thereafter, (2) the orchestrator may receive the edge data and select a control signal to be used in the device 100 from among a plurality of control signals based on the received edge data. Here, the control signal is a control method to be performed by the device 100 and may be a signal that controls the generation of driving control data. For example, the control signal may be a signal instructing to generate driving control data using the first auto encoder 120 or a signal instructing generating driving control data using the second auto encoder 130. .

Meanwhile, even when the orchestrator does not receive edge data, the orchestrator may select a control signal to be used in the device 100 from among a plurality of control signals.

Afterwards, (3) the device 100 can use the control signal generated by the orchestrator. At this time, as shown in FIG. 6A, when the edge node 10 includes the orchestrator 30, the device 100 transmits the control signal generated from the orchestrator 30 included in the edge node 10. You can receive it.

Thereafter, (4) the device 100 generates data (driving data or driving data and edge data) to be used to generate driving control data according to the control signal generated by the orchestrator and the first auto encoder 120 or the second auto encoder. (130) can be set.

Thereafter, (5) the device 100 may make a final decision on the data (driving data or driving data and edge data) to be used to generate driving control data according to the control signal received from the orchestrator.

Thereafter, (6) the auto encoder (first auto encoder 120 or second auto encoder 130) included in the device 100 may convert the determined data according to the control policy.

Finally, (7) driving control data of the device 100 may be generated based on the output value of the auto encoder (first or second auto encoder) selected according to the control signal generated by the orchestrator.

In more detail, when the control signal generated by the orchestrator indicates that driving control data is generated using the first auto encoder 120, the device 100 generates the first driving control data using the first auto encoder 120. If control data is generated and the generated control signal indicates that driving control data is generated using the second auto encoder 130, the second driving control data may be generated using the second auto encoder 130. You can.

7 is an exemplary flowchart of the procedure of a method for generating driving control data according to an embodiment of the present invention. The driving control data generation method of FIG. 7 can be performed by the device 100 shown in FIG. 2 or the device 100 shown in FIG. 3. In addition, the method of generating driving control data shown in FIG. 7 is merely illustrative.

Referring to FIG. 7, the data acquisition unit 110 may acquire driving data around the device 100 (step S10).

For example, the data acquisition unit 110 may use at least one of a camera and a LiDAR sensor to acquire at least one of LiDAR data and image data regarding an obstacle within a predetermined distance from the device 100.

Thereafter, the driving control data generator 140 may generate driving control data using driving data or using driving data and edge data according to a control signal generated based on edge data obtained from an edge node ( Step S20).

In more detail, the driving control data generator 140 may generate driving control data using either the first auto encoder 120 or the second auto encoder 130 according to the generated control signal.

For example, when the generated control signal indicates that the driving control data is generated using the first auto encoder 120, the driving control data generator 140 uses the first auto encoder 120 to generate the driving control data. 1 When driving control data is generated and the generated control signal indicates that driving control data is generated using the second auto encoder 130, the second driving control data is generated using the second auto encoder 130. can be created.

As discussed above, according to the device according to the embodiment of the present invention, an appropriate control signal is generated even in an environment where communication between the edge node and the device is not smooth or is limited, and the driving data obtained from the device is transmitted according to the generated control signal. Alternatively, driving control data can be generated using driving data and edge data obtained from an edge node.

Additionally, devices according to embodiments of the present invention can be used in robots, autonomous drones, and smart factories that operate based on edge computing.

Combinations of each block of the block diagram and each step of the flow diagram attached to the present invention may be performed by computer program instructions. Since these computer program instructions can be mounted on the encoding processor of a general-purpose computer, special-purpose computer, or other programmable data processing equipment, the instructions performed through the encoding processor of the computer or other programmable data processing equipment are included in each block or block of the block diagram. Each step of the flowchart creates a means to perform the functions described. These computer program instructions may also be stored in computer-usable or computer-readable memory that can be directed to a computer or other programmable data processing equipment to implement a function in a particular way, so that the computer-usable or computer-readable memory The instructions stored in can also produce manufactured items containing instruction means that perform the functions described in each block of the block diagram or each step of the flow diagram. Computer program instructions can also be mounted on a computer or other programmable data processing equipment, so that a series of operational steps are performed on the computer or other programmable data processing equipment to create a process that is executed by the computer, thereby generating a process that is executed by the computer or other programmable data processing equipment. Instructions that perform processing equipment may also provide steps for executing functions described in each block of the block diagram and each step of the flow diagram.

Additionally, each block or each step may represent a module, segment, or portion of code that includes one or more executable instructions for executing specified logical function(s). Additionally, it should be noted that in some alternative embodiments it is possible for the functions mentioned in the blocks or steps to occur out of order. For example, two blocks or steps shown in succession may in fact be performed substantially simultaneously, or the blocks or steps may sometimes be performed in reverse order depending on the corresponding function.

The above description is merely an illustrative explanation of the technical idea of the present invention, and those skilled in the art will be able to make various modifications and variations without departing from the essential quality of the present invention. Accordingly, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but are for illustrative purposes, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention shall be interpreted in accordance with the claims below, and all technical ideas within the scope equivalent thereto shall be construed as being included in the scope of rights of the present invention.

10: Edge node
100: device
110: Data acquisition unit
120: first autoencoder
130: second auto encoder
140: Driving control data generation unit

Claims (18)

In an edge computing environment, a device that communicates with an edge node and generates driving control data,
a data acquisition unit that acquires driving data around the device; and
Driving to generate driving control data for autonomous driving of the device by selecting whether to use the driving data or using the driving data and the edge data according to a control signal generated based on edge data obtained from the edge node. Includes a control data generation unit,
The edge data is data measured using a radar included in the edge node,
The driving data is data about obstacles around the device,
Device. According to claim 1,
The control signal is generated by the edge node
Device. According to claim 1,
a first auto-encoder that generates first driving control data using the driving data; and
Further comprising a second auto encoder that generates second driving control data using the driving data and the edge data,
The driving control data generator,
According to the control signal, the driving control data is generated using any one of the first auto encoder and the second auto encoder.
Device. According to claim 3,
The driving control data generator,
When the control signal indicates that the driving control data is generated using the first auto encoder, the first driving control data is generated using the first auto encoder,
When the control signal indicates that the driving control data is generated using the second auto encoder, generating the second driving control data using the second auto encoder
Device. According to claim 3,
The first auto encoder is,
Driving data for learning, and label data for the driving data for learning, are pre-trained to encode the driving data for learning and output the first driving control data when the first correct driving control data is input,
The second auto encoder is,
Upon receiving edge data for learning, driving data for learning, and label data for the edge data for learning and driving data for learning, and driving control data for a second correct answer, the edge data for learning and driving data for learning are encoded to provide the second correct answer. Pre-trained to output driving control data
Device. According to claim 3,
Further comprising an orchestrator that receives the edge data from the edge node and generates the control signal.
Device. According to claim 6,
The orchestrator is,
When learning edge data is input, it is pre-trained to output a learning control signal for selecting one of the first auto encoder and the second auto encoder,
By using a loss function for the difference between the learning control signal and the correct answer control signal, when the learning edge data is input, the learning control signal is further learned to be output.
Device. According to claim 1,
The data acquisition unit,
Obtaining the driving data using at least one of a camera and a lidar sensor
Device. In an edge computing environment, a method of generating driving control data performed by a device that communicates with an edge node and generates driving control data, comprising:
acquiring driving data around the device;
Generating driving control data for autonomous driving of the device by selecting whether to use the driving data or using the driving data and the edge data according to a control signal generated based on edge data obtained from the edge node. Including,
The edge data is data measured using a radar included in the edge node,
The driving data is data about obstacles around the device,
How to generate driving control data. According to clause 9,
The control signal is generated by the edge node
How to generate driving control data. According to clause 9,
The step of generating the driving control data is,
According to the control signal, either a first autoencoder that generates first travel control data using the driving data and a second autoencoder that generates second driving control data using the driving data and the edge data. Generating the driving control data using
How to generate driving control data. According to claim 11,
When the control signal indicates that the driving control data is generated using the first auto encoder, the first driving control data is generated using the first auto encoder,
When the control signal indicates that the driving control data is generated using the second auto encoder, generating the second driving control data using the second auto encoder
How to generate driving control data. According to claim 11,
The first auto encoder is,
Driving data for learning, and label data for the driving data for learning, are pre-trained to encode the driving data for learning and output the first driving control data when the first correct driving control data is input,
The second auto encoder is,
Upon receiving edge data for learning, driving data for learning, and label data for the edge data for learning and driving data for learning, and driving control data for a second correct answer, the edge data for learning and driving data for learning are encoded to provide the second correct answer. Pre-trained to output driving control data
How to generate driving control data. According to claim 11,
Further comprising receiving edge data from the edge node and generating the control signal.
How to generate driving control data. According to claim 14,
The step of generating the control signal is,
When learning edge data is input, it is pre-trained to output a learning control signal for selecting one of the first auto encoder and the second auto encoder,
Using a loss function for the difference between the learning control signal and the correct answer control signal, when receiving the learning edge data, the control signal is generated using an orchestrator that is further learned to output the learning control signal.
How to generate driving control data. According to clause 9,
The step of acquiring the driving data is,
Obtaining the driving data using at least one of a camera and a lidar sensor
How to generate driving control data. A computer-readable recording medium storing a computer program,
When the computer program is executed by a processor,
In an edge computing environment, acquiring driving data surrounding a device that communicates with an edge node and generates driving control data;
According to a control signal generated based on edge data obtained from the edge node, using the driving data or selecting whether to use the driving data and the edge data to generate driving control data for autonomous driving of the device Includes instructions for causing the processor to perform a method including the steps,
The edge data is data measured using a radar included in the edge node,
The driving data is data about obstacles around the device,
A computer-readable recording medium. A computer program stored on a computer-readable recording medium,
When the computer program is executed by a processor,
In an edge computing environment, acquiring driving data surrounding a device that communicates with an edge node and generates driving control data;
According to a control signal generated based on edge data obtained from the edge node, using the driving data or selecting whether to use the driving data and the edge data to generate driving control data for autonomous driving of the device Includes instructions for causing the processor to perform a method including the steps,
The edge data is data measured using a radar included in the edge node,
The driving data is data about obstacles around the device,
computer program.

Images