System
Through a multi-level proxy structure, the problem of low information processing efficiency in the existing technology is solved, and efficient information analysis and processing in cloud, fog, and edge computing environments is realized. It is suitable for IoT devices Information analysis and autonomous driving applications.
Patent Information
- Application Number
- CN202380033774.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-04-14
- Filing Date
- 2023-04-06
- Publication Date
- 2024-12-06
AI Technical Summary
The prior art is difficult to effectively collaborate and process information in multi-agent learning systems, resulting in low information processing efficiency under the network hierarchy structure and cannot meet the information analysis and processing needs of IoT devices in cloud, fog, and edge computing environments.
Adopting a multi-level proxy structure, through collaboration between superior agents, intermediate agents and lower agents, information collection, learning execution and information transmission are carried out separately, and hierarchical processing and collaboration of information are realized, and cloud, fog, and edge computing resources are used for real-time and non-real-time Analysis, optimize information routing and network control.
It improves information processing efficiency and network stability, adapts to a large amount of information sent by IoT devices, realizes real-time and robust information analysis and processing, and is suitable for application scenarios such as accident prediction and autonomous driving.
Smart Images

Figure CN119096254A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a system. Background Art
[0002] Patent Document 1 describes multi-agent reinforcement learning.
[0003] Prior art literature
[0004] Patent Literature
[0005] Patent Document 1: Japanese Patent Application Publication No. 2004-227208 Summary of the invention
[0006] According to a first aspect of the present invention, a system is provided. The system may include a first-level agent configured at a first level of a network and a plurality of second-level agents at a second level lower than the first level. Each of the plurality of second-level agents may perform learning by using the collected information in cooperation with other second-level agents, and send information generated using the learning result to the first-level agent. The first-level agent may perform learning by using the information received from the plurality of second-level agents. The system may further include a plurality of third-level agents at a third level lower than the second level. Each of the plurality of third-level agents may perform learning by using the collected information in cooperation with other third-level agents, and send information generated using the learning result to at least any one of the plurality of second-level agents. Each of the plurality of third-level agents may include: a third information collection unit that collects information; a third learning execution unit that performs learning by using the information collected by the third information collection unit in cooperation with other plurality of third-level agents; and a third information transmission unit that sends information generated using the learning result of the third learning execution unit to at least any one of the plurality of second-level agents. Each of the plurality of second-level agents may include: a second information collection unit that collects information from the plurality of third-level agents; a second learning execution unit that uses the information collected by the second information collection unit to perform learning in cooperation with other plurality of second-level agents; and a second information transmission unit that transmits information generated using the learning result of the second learning execution unit to the first-level agent. The first-level agent may include: a first information collection unit that collects information from the plurality of second-level agents; and a first learning execution unit that performs learning using the information collected by the first information collection unit. The first-level agent may be configured at a first NW level of a hierarchical network, the plurality of second-level agents may be configured at a second NW level lower than the first NW level of the hierarchical network, and the plurality of third-level agents may be configured at a third NW level lower than the second NW level of the hierarchical network. The hierarchical network may be a cloud network, the first NW level may be composed of cloud computing, the second NW level may be composed of a plurality of fog computing, and the third NW level may be composed of a plurality of edge computing.The hierarchical network may be a cloud network, the first NW layer may be composed of cloud computing, the second NW layer may be composed of multiple fog computing, the third NW layer may be composed of multiple edge computing, the third information collection unit may collect information from multiple IoT (Internet of Thing) devices through mobile communication, the third information sending unit may use the learning result of the third learning execution unit to send information selected from the information collected by the third information collection unit, or information generated using the learning result, to at least any one of the multiple second-level agents, the second information collection unit may collect information sent by the third information sending unit, and the second information sending unit may use the learning result of the second learning execution unit to send information selected from the information collected by the second information collection unit, or information generated using the learning result, to the first-level agent.
[0007] In the system, each of the plurality of third-level agents may include a third processing execution unit, and the third processing execution unit performs processing on the information collected by the third information collection unit using the learning result of the third learning execution unit. The third processing execution unit may generate information to be sent to at least any one of the plurality of second-level agents using the plurality of information collected by the third information collection unit. The third processing execution unit may generate transmission information including information selected from the plurality of information collected by the third information collection unit using the learning result of the third learning execution unit. The third processing execution unit may generate summary information that summarizes the plurality of information collected by the third information collection unit using the learning result of the third learning execution unit. The third information sending unit may send the information generated by the third processing execution unit to at least any one of the plurality of second-level agents. In the system, each of the plurality of second-level agents may include a second processing execution unit, and the second processing execution unit performs processing on the information collected by the second information collection unit using the learning result of the second learning execution unit. The second processing execution unit may generate information to be sent to the first-level agent using the plurality of information collected by the second information collection unit. The second processing execution unit may generate transmission information including information selected from a plurality of information collected by the second information collection unit using the learning result of the second learning execution unit. The second processing execution unit may generate summary information that summarizes a plurality of information collected by the second information collection unit using the learning result of the second learning execution unit. The second information transmission unit may transmit the information generated by the second processing execution unit to the first hierarchical agent. In the system, the first hierarchical agent may have a first processing execution unit that performs processing on the information collected by the first information collection unit using the learning result of the first learning execution unit. The first processing execution unit may perform processing for the purpose of stabilizing the entire network to which the system is applied using the learning result of the first learning execution unit. The first processing execution unit may generate instruction information for at least any one of the plurality of second hierarchical agents or at least any one of the plurality of third hierarchical agents based on the result of analyzing the information collected by the first information collection unit. The first processing execution unit may generate instruction information that instructs at least any one of the plurality of third hierarchical agents to transmit information collected from the IoT device to at least any one of the plurality of second hierarchical agents. The first processing execution unit may generate instruction information that instructs at least one of the plurality of second-level agents on processing content of information collected from at least one of the plurality of second-level agents.The first processing execution unit may generate learning information to be sent to at least any one of the plurality of third-level agents based on the result of analyzing the information collected by the first information collection unit. The first processing execution unit may generate learning information to be sent to at least any one of the plurality of second-level agents based on the result of analyzing the information collected by the first information collection unit. The first processing execution unit may generate learning information including a reward set in accordance with a change in the tendency of the information collected by the first information collection unit.
[0008] In the case where the system aims to analyze information sent by a plurality of IoT devices, the plurality of third-layer agents may perform analysis that further requires real-time performance, the first-layer agent analyzes a tendency of a certain long period, and the plurality of second-layer agents perform analysis corresponding to the period. In the case where the system aims to manage the occurrence of accidents in a certain area, the plurality of third-layer agents may collect image data and object detection data from a plurality of IoT devices arranged in the area, the plurality of third-layer agents in charge of geographically close sub-areas may perform learning for detecting the occurrence of accidents while sharing information, the third-layer agents may detect the occurrence of accidents using the learning results, the plurality of second-layer agents may be assigned to a group of sub-areas, and the third-layer agents may collect information from the IoT devices of the sub-areas of the group, the plurality of second-layer agents may perform learning for predicting the occurrence of accidents by cooperating with each other, the second-layer agents may predict the occurrence of accidents using the learning results, and the first-layer agents may perform learning for controlling the overall information based on the prediction results of the plurality of second-layer agents. The first-level agent can perform the following control: for the sub-area where the accident is predicted to occur, the amount of information collected or the types of information can be increased; for the other sub-areas, the amount of information collected or the types of information can be reduced.
[0009] For example, in the case where the system has the purpose of smoothly implementing the routing of messages notifying information acquired by IoT devices in the cloud network, the third-level proxy can implement the routing of messages. The third-level proxy can cooperate with other third-level proxies to perform learning to control Topic, To (including Copy) and segmentation for messages from publishers. The third-level proxy can use the learning results to cooperate with other third-level proxies to control the generation of Topic, determination of To, copying of messages, segmentation of messages, etc., so that messages from publishers reach appropriate subscribers. The second-level proxy can, for example, monitor the routing based on the third-level proxy based on the information collected from the third-level proxy and perform learning that can perform processing to eliminate problems generated in routing. The second-level proxy can buffer messages when there are messages with the same To whose destination is unknown, and release part of the buffered messages to the network when the buffered amount exceeds a threshold. As a result, in the case of returning again, buffer until arrival, or buffer for a certain period of time, and if messages start to arrive, start sending messages with the same To in the buffered messages. The first-level agent may perform learning for detecting a problem point in the network based on information collected from the second-level agent. When a problem point in the network is detected using the learning result, the first-level agent may notify the operator of the network of the detection result or output an instruction to change the network structure.
[0010] For example, in the case where the system has the purpose of implementing control related to autonomous driving, each of the plurality of third-level agents can collect information from IoT devices mounted on vehicles located in the region, IoT devices mounted on traffic lights, etc., and IoT devices installed on roads, etc., for each region divided into a plurality of regions. The third-level agent can collect at least one of the vehicle's location information, images captured by a vehicle camera or a street camera, vehicle detection information detected by a road sensor, human detection information, the distance between vehicles detected by a vehicle sensor, etc., weather information in the region, navigation information of the vehicle, and the driving speed of the vehicle. The third-level agent can be responsible for a certain area, and when a certain vehicle enters the area, it collects information from the IoT device mounted on the vehicle, and when the vehicle leaves the area, the third-level agent responsible for the area entered by the vehicle takes over the collection of information. The third-level lower-level agent can share information with other third-level agents, and perform detection of the danger of the vehicle-to-vehicle distance being shorter than a threshold and the vehicle colliding with a person, etc. When a danger is detected, the third-level agent can send danger detection information to the target vehicle. The second-tier agent can instruct the third-tier agent what information to communicate. The second-tier agent can analyze (learn) important information (many people, many cars, etc.) based on the information (day of the week, time, etc.) collected from the third-tier agent in the past, and can determine the information with high priority to be collected. The second-tier agent can issue an instruction to the third-tier agent so that the third-tier agent can collect high-priority information for each period. The second-tier agent can learn by cooperating with other second-tier agents and can determine high-priority information with high accuracy. As a result of learning, the second-tier agent can instruct in the morning of weekdays so that the priority of car detection results can be increased to collect information, and instruct in the daytime of holidays so that the priority of human detection results can be increased to collect information. The first-tier agent can perform a wider range of analysis. The first-tier agent can perform a forecast (at a set time) after several steps based on the traffic information (traffic volume, time, day of the week, event, weather) of the responsible area (city, county, national unit), and provide information to the second-tier agent.
[0011] In addition, the above summary of the invention does not list all the necessary features of the present invention. In addition, sub-combinations of these feature groups may also constitute inventions. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] Figure 1 An example of a network 100 to which the autonomous distributed system is applied is schematically shown.
[0013] Figure 2An example of a cloud network 300 is schematically shown as an example of the network 100 .
[0014] Figure 3 An example of an autonomous distributed system 200 is schematically shown.
[0015] Figure 4 An example of the functional configuration of the lower agent 230 is schematically shown.
[0016] Figure 5 An example of the functional configuration of the intermediate agent 220 is schematically shown.
[0017] Figure 6 An example of the functional configuration of the host agent 210 is schematically shown.
[0018] Figure 7 Another example of an autonomous distributed system 200 is schematically shown.
[0019] Figure 8 An example of the hardware configuration of the computer 1200 that realizes the upper agent 210 , the intermediate agent 220 , or the lower agent 230 is schematically shown. DETAILED DESCRIPTION
[0020] An autonomous distributed system is known as one of the multi-agent systems. In an autonomous distributed system, the purpose is for each of the multiple agents to learn independently and autonomously to obtain cooperative actions. The completely identical state (structure or parameter) of each agent is called homogenious, and the state that changes from homogenious to different states (heterogenious) is called functional differentiation. As an application example of an autonomous distributed system, there is an example of the following tracking problem: there is a target and multiple hunters, and multiple hunters learn the method of capturing the target, that is, multiple hunter agents learn their own roles respectively, and collaborate to capture the target. In an autonomous distributed system, a group of agents with the same performance (homo) is functionally differentiated by learning, and each has a role (hetero). For example, in a tracking problem, it is divided into a direct pursuit agent and a roundabout agent. By using such a technology, for example, it is possible to autonomously learn to adapt to the environment through the same algorithm, or to learn autonomously and distributedly through multiple agents. In this embodiment, an autonomous distributed system is applied to a network.
[0021] The present invention will be described below by way of the embodiments of the invention, but the following embodiments do not limit the invention according to the claims. In addition, not all combinations of features described in the embodiments are essential to the solution means of the invention.
[0022] Figure 1An example of a communication network 100 to which the autonomous distributed system is applied is schematically shown. The communication network 100 is a hierarchical network and is composed of an upper-level calculation 110 , a plurality of intermediate calculations 120 , and a plurality of lower-level calculations 130 .
[0023] The upper calculation 110 constitutes a first NW (Network) layer, the middle calculations 120 constitute a second NW layer lower than the first NW layer, and the lower calculations 130 constitute a third NW layer lower than the second NW layer.
[0024] The communication network 100 to which the autonomous distributed system is applied may be any communication network as long as it has a hierarchical structure. For example, the communication network 100 is a cloud network.
[0025] Figure 2 An example of a cloud network 300 is schematically shown as an example of the communication network 100. The cloud network 300 is composed of a cloud computing 310, a plurality of fog computing 320, and a plurality of edge computing 330.
[0026] Cloud computing 310 may be an example of upper computing 110 . Fog computing 320 may be an example of intermediate computing 120 . Edge computing 330 may be an example of lower computing 130 .
[0027] In the cloud network 300 , the first NW level is composed of cloud computing 310 , the second NW level is composed of multiple fog computing 320 , and the third NW level is composed of multiple edge computing 330 .
[0028] Each of the plurality of edge computing devices 330 communicates with one or more IoT devices 400 via a mobile communication system. Each of the plurality of IoT devices 400 can send information to at least one of the plurality of edge computing devices 330 via a wireless base station and a Wi-Fi (registered trademark) access point.
[0029] The IoT device 400 may be any device as long as it can obtain or send any information. The IoT device 400 may be equipped with various sensors, for example. Examples of information sent by the IoT device 400 include image data (still images, dynamic images), sound data, infrared data, location data, object detection data, distance data, weather data, temperature data, and humidity data, but these are only examples and may be any information.
[0030] The mobile communication system is, for example, a 5G (5th Generation) communication system. The mobile communication system may also be an LTE (Long Term Evolution) communication system. The mobile communication system may also be a 3G (3rd Generation) communication system. The mobile communication system may also be a mobile communication system after a 6G (6th Generation) communication system.
[0031] Figure 3 An example of an autonomous distributed system 200 is schematically shown. The autonomous distributed system 200 is composed of a plurality of agents having a hierarchical structure. Figure 3 In the illustrated example, the autonomous distributed system 200 includes an upper agent 210 , a plurality of intermediate agents 220 , and a plurality of lower agents 230 .
[0032] The autonomous distribution system 200 is applied to the communication network 100. For example, the upper agent 210 is configured in the upper computing 110, each of the plurality of intermediate agents 220 is configured in each of the plurality of intermediate computing 120, and each of the plurality of lower agents 230 is configured in each of the plurality of lower computing 130. It is also possible to configure a plurality of intermediate agents 220 for one intermediate computing 120. It is also possible to configure a plurality of lower agents 230 for one lower computing 130.
[0033] exist Figure 2 , the case where the autonomous distributed system 200 is composed of three levels is illustrated, but it is not limited to this. The autonomous distributed system 200 may be two levels, or may be four levels or more. In the case where the autonomous distributed system 200 is two levels, the autonomous distributed system 200 may include an upper agent 210 and multiple lower agents 230. In this example, the upper agent 210 may be an example of a first-level agent of the first level, and the lower agent 230 may be an example of multiple second-level agents of the second level. In this case, for example, the upper agent 210 is configured in the upper calculation 110, and each of the multiple lower agents 230 is configured in each of the multiple lower calculations 130. In addition, for example, the upper agent 210 is configured in any one of the multiple intermediate calculations 120, and the multiple lower agents 230 are configured in each of the multiple lower calculations 130. In addition, in this case, the communication network 100 may also be composed of two levels without multiple intermediate calculations 120.
[0034] The autonomous distribution system 200 is applied to a cloud network 300, for example. For example, the upper agent 210 is configured in the cloud computing 310, each of the plurality of intermediate agents 220 is configured in each of the plurality of fog computing 320, and each of the plurality of lower agents 230 is configured in each of the plurality of edge computing 330. Multiple intermediate agents 220 may be configured for one fog computing 320. Multiple lower agents 230 may also be configured for one edge computing 330.
[0035] In the case where the autonomous distributed system 200 is a two-level system, for example, the upper agent 210 is configured in the cloud computing 310, and each of the plurality of lower agents 230 is configured in each of the plurality of edge computing 330. In addition, for example, the upper agent 210 is configured in any one of the plurality of fog computing 320, and the plurality of lower agents 230 is configured in each of the plurality of edge computing 330. In addition, in this case, the cloud network 300 may not include the plurality of fog computing 320 but may be composed of two levels.
[0036] The lower agent 230 collects information. The lower agent 230, for example, collects information sent by the IoT device 400. The lower agent 230 may collect information from the IoT device 400 through mobile communication. The lower agent 230 may also collect information sent by any device other than the IoT device 400. The lower agent 230 may collect information from any device other than the IoT device 400 through mobile communication.
[0037] The lower-level agent 230 uses the collected information to perform learning in cooperation with other lower-level agents 230. The lower-level agent 230 can perform learning according to the pre-registered remuneration. The lower-level agent 230 can perform learning according to the pre-registered knowledge. The lower-level agent 230 can use the learning results to perform various processes. For example, the lower-level agent 230 uses the learning results to select information to be sent to the intermediate agent 220 from the collected multiple information and sends it to the intermediate agent 220. For example, the lower-level agent 230 uses the learning results to generate summary information that summarizes the collected multiple information and sends it to the intermediate agent 220.
[0038] For example, in the case where the purpose is to send only useful information from a large amount of information sent by multiple IoT devices 400 to the upper level, the usefulness of the information collected from the IoT device 400 and sent to the intermediate agent 220 or the upper level agent 210 by the lower level agent 230 is registered as a reward. Each of the multiple lower level agents 230 cooperates with other lower level agents 230 while independently learning to maximize the reward. The multiple lower level agents 230 can collect information from different one or more IoT devices 400, select information to be sent to the upper level from the collected information based on different criteria, or generate summary information to be sent to the upper level from the collected information, and construct the system as a whole so that only information with high usefulness is transmitted to the upper level.
[0039] The lower-level agent 230 can perform processing with relatively high real-time performance. For example, the lower-level agent 230 uses the learning results to select and process the information collected from one or more IoT devices 400 at predetermined times such as every 1 minute and 3 minutes, or generates summary information, and sends the information to the upper level. As a result, the information sent to the upper level can be adjusted in real time. For example, in the case of multiple lower-level agents 230 configured in the edge computing 330 that processes more images during the day and more infrared data at night, multiple lower-level agents 230 are responsible for images during the day and infrared data at night. In this way, the lower-level agent 230 can dynamically adapt to the environment, so even if the IoT device 400 of the collection object is changed, or the type of information collected by the IoT device 400 of the collection object changes, it can adapt. In this way, the lower-level agent 230 can perform processing in real time, but it can also send statistical information or send summary information in a time-delayed manner. It is also possible that the lower-level agent 230 can intentionally delay.
[0040] The intermediate agent 220 collects information. The intermediate agent 220 may collect information from the lower agent 230. The intermediate agent 220, for example, collects information collected by the lower agent 230 from the IoT device 400 from the lower agent 230. The intermediate agent 220, for example, collects information selected by the lower agent 230 from the plurality of information collected by the lower agent 230 from the IoT device 400 from the lower agent 230. The intermediate agent 220, for example, collects summary information that summarizes the plurality of information collected by the lower agent 230 from the IoT device 400 from the lower agent 230.
[0041] The intermediate agent 220 uses the collected information to perform learning in cooperation with other intermediate agents 220. The intermediate agent 220 can perform various processes using the learning results. For example, the intermediate agent 220 uses the learning results to select information to be sent to the intermediate agent 220 from among the collected multiple information and sends it to the upper agent 210. For example, the intermediate agent 220 uses the learning results to generate summary information that summarizes the collected multiple information and sends it to the upper agent 210.
[0042] The intermediate agent 220 can perform processing with lower real-time performance than the lower agent 230. For example, the intermediate agent 220 performs processing using information collected from one or more lower agents 230 at predetermined periods such as every hour and every day. The intermediate agent 220, for example, performs learning using information of a predetermined period. The intermediate agent 220 can also cooperate with other intermediate agents 220 to adjust the amount of information sent from multiple lower agents 230 to the upper level, or adjust the amount of information sent from multiple intermediate agents 220 to the upper agent 210. The intermediate agent 220 can also perform learning and processing to ensure robustness in the autonomous distribution system 200. The intermediate agent 220 can also play the role of the Spinal Cord in the autonomous distribution system 200. When receiving information from the lower agent 230, the intermediate agent 220 can also determine whether to send the information or information generated based on the information to the upper agent 210, or to respond to or send instructions to the lower agent 230.
[0043] The upper agent 210 collects information. The upper agent 210 may collect information from the middle agent 220. The upper agent 210 collects, for example, from the lower agent 230 the information that the middle agent 220 has collected from the lower agent 230. The upper agent 210 collects, for example, from the middle agent 220 information selected by the middle agent 220 from among a plurality of information that the middle agent 220 has collected from the lower agent 230. The upper agent 210 collects, for example, from the middle agent 220 summary information that summarizes a plurality of information that the middle agent 220 has collected from the lower agent 230. The upper agent 210 may also collect information from the lower agent 230.
[0044] The upper agent 210 executes processing using the collected information. For example, the upper agent 210 executes learning using the collected information. For example, the upper agent 210 executes various processing using the learning result.
[0045] The upper agent 210 may perform processing with lower real-time performance than the intermediate agent 220. For example, the upper agent 210 performs processing using collected information at predetermined intervals such as every week, month, and year. The upper agent 210 performs learning using information of a predetermined period, for example.
[0046] The upper agent 210 may also perform processing for the purpose of stabilizing the overall autonomous distribution system 200. For example, the upper agent 210 may perform a major cause analysis of the overall communication load of the communication network 100 and the cloud network 300 based on the collected information. The upper agent 210 may determine the location where the communication load in the network has increased based on the content of the collected information and the collection status of the information, such as adjusting the amount of information sent by the IoT device 400, the lower agent 230, and the intermediate agent 220, or adjusting the transmission path to reduce the communication load.
[0047] The amount of information processed in the so-called cloud network has increased significantly. In the future, the popularity of 5G mobile communication systems will further develop, and it is expected that the amount of information will further increase due to the increase in the number of IoT devices. For such an increase in the amount of information, improving the communication function of the network and improving the performance of network equipment have become conventional countermeasures, but sometimes it is not possible to fully cope with it only through this countermeasure. In contrast, by applying the autonomous distribution system 200 involved in this embodiment to the network, the processing can be hierarchical. According to the autonomous distribution system 200, for example, while performing high-real-time processing through the lower agent 230, it is reduced to useful information and passed to the upper level, while performing slightly wider span information processing using strictly selected information through the intermediate agent 220, the information is further reduced, and in the upper agent 210, a wider span information processing using more strictly selected information is performed, or the overall stabilization of the network is achieved, thereby helping to build an environment that can properly process a large amount of information.
[0048] Figure 4 An example of the functional configuration of the lower agent 230 is schematically shown. The lower agent 230 includes a storage unit 231 , a registration unit 232 , an information collection unit 233 , a learning execution unit 234 , a process execution unit 235 , and an information transmission unit 236 .
[0049] The registration unit 232 performs various registrations. The registration unit 232, for example, registers the learning information used by the lower agent 230 for learning. The registration unit 232 stores the registered learning information in the storage unit 231. The learning information may include a reward. The learning information may include knowledge. The registration unit 232, for example, registers the learning information by receiving input from an operator of the autonomous distribution system 200. The registration unit 232, for example, registers the learning information received from the upper agent 210. The registration unit 232, for example, registers the learning information received from the intermediate agent 220. The registration unit 232 may be an example of a third registration unit.
[0050] The information collecting unit 233 collects information. The information collecting unit 233 may collect information sent by any device. The information collecting unit 233, for example, collects information sent by the IoT device 400. The information collecting unit 233 may collect information from the IoT device 400 through mobile communication. The information collecting unit 233 stores the collected information in the storage unit 231. The information collecting unit 233 may be an example of a third information collecting unit.
[0051] The learning execution unit 234 executes learning using the information collected by the information collection unit 233. The learning execution unit 234 can execute learning by cooperating with other lower agents 230 using the information collected by the information collection unit 233.
[0052] The learning execution unit 234 stores the learning result in the storage unit 231. The learning result may include a model generated by the learning execution unit 234. The learning result may include a neural network generated by the learning execution unit 234. The learning execution unit 234 may be an example of a third learning execution unit.
[0053] The learning execution unit 234 can use any learning method. For example, the learning execution unit 234 first performs simulation-based pre-learning (pre-training), and updates the model and neural network through the information collected by the information collection unit 233. In pre-learning, actual data collected by the information collection unit 233 in the past can be used. Synthetic data can also be used in pre-learning. In the case of an environment where trial and error is not allowed during learning, it is effective to learn in a rule base instead of reinforcement learning, and only learn its parameters. The learning method can also be ANN (Artificial Neural Network), DNN (Deep Neural Network), heuristic method, etc.
[0054] The process execution unit 235 executes various processes. The process execution unit 235 may execute processes on the information collected by the information collection unit 233 using the learning results of the learning execution unit 234. The process execution unit 235 may generate information to be sent to the intermediate agent 220 using the plurality of information collected by the information collection unit 233.
[0055] For example, the processing execution unit 235 generates transmission information including information selected from a plurality of information collected by the information collection unit 233 using the learning result of the learning execution unit 234. For example, the processing execution unit 235 generates summary information that summarizes a plurality of information collected by the information collection unit 233 using the learning result of the learning execution unit 234. The processing execution unit 235 may be an example of a third processing execution unit.
[0056] The information transmission unit 236 transmits information. For example, the information transmission unit 236 transmits information collected by the information collection unit 233 to the intermediate agent 220. For example, the information transmission unit 236 transmits information generated by the process execution unit 235 to the intermediate agent 220. The information transmission unit 236 may be an example of a third information transmission unit.
[0057] Figure 5 An example of the functional configuration of the intermediate agent 220 is schematically shown. The intermediate agent 220 includes a storage unit 221 , a registration unit 222 , an information collection unit 223 , a learning execution unit 224 , a process execution unit 225 , and an information transmission unit 226 .
[0058] The registration unit 222 performs various registrations. The registration unit 222, for example, registers the learning information used by the intermediate agent 220 for learning. The registration unit 222 stores the registered learning information in the storage unit 221. The learning information may include a reward. The learning information may include knowledge. The registration unit 222, for example, registers the learning information by receiving input from an operator of the autonomous distribution system 200. The registration unit 222, for example, registers the learning information received from the upper agent 210. The registration unit 222 may be an example of a second registration unit.
[0059] The information collecting unit 223 collects information. The information collecting unit 223 may collect information from the lower agent 230. The information collecting unit 223 collects information transmitted by the information transmitting unit 236 of the lower agent 230, for example. The information collecting unit 223 stores the collected information in the storage unit 221. The information collecting unit 223 may be an example of a second information collecting unit.
[0060] The learning execution unit 224 executes learning using the information collected by the information collection unit 223. The learning execution unit 224 may execute learning in cooperation with other intermediate agents 220 using the information collected by the information collection unit 223.
[0061] The learning execution unit 224 stores the learning result in the storage unit 221. The learning result may include a model generated by the learning execution unit 224. The learning result may include a neural network generated by the learning execution unit 224. The learning execution unit 224 may be an example of a second learning execution unit.
[0062] The learning execution unit 224 can use any learning method. For example, the learning execution unit 224 first performs simulation-based pre-learning, and updates the model and neural network through the information collected by the information collection unit 223. In the pre-learning, actual data collected in the past by the information collection unit 223 can be used. Synthetic data can also be used in pre-learning. In the case of an environment where trial and error is not allowed during learning, it is effective to learn in a rule base instead of reinforcement learning, and only learn its parameters. The learning method can also be ANN, DNN, heuristic method, etc.
[0063] The processing execution unit 225 executes various processes. The processing execution unit 225 can execute processes on the information collected by the information collection unit 223 using the learning results of the learning execution unit 224. The processing execution unit 225 can generate information to be transmitted to the upper agent 210 using the plurality of information collected by the information collection unit 223.
[0064] For example, the processing execution unit 225 generates transmission information including information selected from a plurality of information collected by the information collection unit 223 using the learning result of the learning execution unit 224. For example, the processing execution unit 225 generates summary information that summarizes a plurality of information collected by the information collection unit 223 using the learning result of the learning execution unit 224. The processing execution unit 225 may be an example of a second processing execution unit.
[0065] The information transmission unit 226 transmits information. For example, the information transmission unit 226 transmits information collected by the information collection unit 223 to the upper agent 210. For example, the information transmission unit 226 transmits information generated by the process execution unit 225 to the upper agent 210. The information transmission unit 226 may be an example of a second information transmission unit.
[0066] Figure 6 An example of the functional configuration of the host agent 210 is schematically shown. The host agent 210 includes a storage unit 211 , a registration unit 212 , an information collection unit 213 , a learning execution unit 214 , a process execution unit 215 , and an information transmission unit 216 .
[0067] The registration unit 212 performs various registrations. The registration unit 212, for example, registers the learning information used by the upper agent 210 for learning. The registration unit 212 stores the registered learning information in the storage unit 211. The learning information may include rewards. The learning information may include knowledge. The registration unit 212 registers the learning information by, for example, receiving input from an operator of the autonomous distribution system 200. The registration unit 212 may be an example of a first registration unit.
[0068] The information collecting unit 213 collects information. The information collecting unit 213 may collect information from the intermediate agent 220. The information collecting unit 213 collects information sent by the information sending unit 226 of the intermediate agent 220, for example. The information collecting unit 213 stores the collected information in the storage unit 211. The information collecting unit 213 may be an example of a first information collecting unit.
[0069] The learning execution unit 214 uses the information collected by the information collection unit 213 to perform learning. The learning execution unit 214 may be an example of a first learning execution unit. The learning execution unit 214 may also use the information collected by the information collection unit 213 to cooperate with multiple intermediate agents 220 to perform learning. The learning execution unit 214 may also use the information collected by the information collection unit 213 to cooperate with multiple lower-level agents 230 to perform learning. The learning execution unit 214 may also use the information collected by the information collection unit 213 to cooperate with multiple intermediate agents 220 and multiple lower-level agents 230 to perform learning.
[0070] The learning execution unit 214 stores the learning result in the storage unit 211. The learning result may include a model generated by the learning execution unit 214. The learning result may include a neural network generated by the learning execution unit 214. The learning execution unit 214 may be an example of a first learning execution unit.
[0071] The learning execution unit 214 can use any learning method. For example, the learning execution unit 214 first performs pre-learning based on simulation, and updates the model and neural network through the information collected by the information collection unit 213. In the pre-learning, actual data collected in the past by the information collection unit 213 can be used. Synthetic data can also be used in pre-learning. In the case of an environment where trial and error is not allowed during learning, it is effective to learn in a rule base instead of reinforcement learning and only learn its parameters. The learning method can also be ANN, DNN, heuristic method, etc.
[0072] The processing execution unit 215 executes various processes. The processing execution unit 215 may be an example of a first processing execution unit. The processing execution unit 215 may execute a process on the information collected by the information collection unit 213 using the learning result of the learning execution unit 214. The processing execution unit 215 may execute a process for the purpose of stabilizing the entire network to which the autonomous distribution system 200 is applied using the learning result of the learning execution unit 214.
[0073] The processing execution unit 215 may also generate instruction information for the intermediate agent 220 or the lower agent 230 based on the result of analyzing the information collected by the information collection unit 213. For example, the processing execution unit 215 generates instruction information that indicates to the lower agent 230 the information to be sent to the intermediate agent 220 among the information collected from the IoT device 400. For example, the processing execution unit 215 generates instruction information that indicates to the intermediate agent 220 the content of processing of the information collected from the intermediate agent 220.
[0074] The processing execution unit 215 may also generate learning information to be sent to the lower agent 230 based on the result of analyzing the information collected by the information collection unit 213. The processing execution unit 215 may also generate learning information to be sent to the intermediate agent 220 based on the result of analyzing the information collected by the information collection unit 213. For example, the processing execution unit 215 generates learning information including a reward set in accordance with a change in the tendency of the information collected by the information collection unit 213.
[0075] The information transmission unit 216 transmits information. For example, the information transmission unit 216 transmits the instruction information generated by the process execution unit 215 to the lower agent 230. For example, the information transmission unit 216 transmits the instruction information generated by the process execution unit 215 to the intermediate agent 220.
[0076] For example, the information transmission unit 216 transmits the learning information generated by the process execution unit 215 to the lower agent 230. For example, the information transmission unit 216 transmits the learning information generated by the process execution unit 215 to the intermediate agent 220.
[0077] For example, when the autonomous distributed system 200 has the analysis of information sent from a plurality of IoT devices 400 as one purpose, the lower agent 230 performs analysis that further requires real-time performance, the upper agent 210 analyzes trends over a relatively long period, and the intermediate agent 220 performs analysis corresponding to that period.
[0078] As a specific example, in the case where the purpose is to manage the occurrence of accidents in a certain area, multiple lower-level agents 230 collect image data, object detection data, etc. from multiple IoT devices 400 configured in the area. For example, the area is divided into multiple sub-areas, and each of the multiple lower-level agents 230 collects information from the IoT device 400 configured in each of the multiple sub-areas. Then, the multiple lower-level agents 230 responsible for the geographically close sub-areas share information while performing learning for detecting the occurrence of accidents. The lower-level agent 230 can detect the occurrence of accidents using the learning results. In addition, multiple intermediate agents 220 are assigned to groups of sub-areas, and collect information from the lower-level agents 230 that collect information from the IoT devices 400 of the sub-areas of the group. For example, multiple intermediate agents 220 perform learning for predicting the occurrence of accidents by cooperating with each other. The intermediate agent 220 can predict the occurrence of accidents using the learning results. The upper-level agent 210, for example, performs learning for controlling overall information based on the prediction results of the multiple intermediate agents 220. As a specific example, the upper agent 210 performs control such that the amount of information collected or the types of information are increased for sub-regions where accidents are predicted to occur, and the amount of information collected or the types of information are reduced for other sub-regions.
[0079] For example, in the case where the autonomous distribution system 200 has a purpose of smoothly implementing the routing of messages notifying information acquired by the IoT device 400 in the cloud network 300, the lower agent 230 implements the routing of the message. The lower agent 230 cooperates with other lower agents 230 to perform learning to control Topic, To (including Copy), and segmentation for messages from publishers. The lower agent 230 uses the learning results and cooperates with other lower agents 230 to control the generation of Topic, the determination of To, the copying of messages, the segmentation of messages, etc., so that the messages from the publishers reach the appropriate subscribers.
[0080] The intermediate proxy 220 monitors the routing based on the lower proxy 230, for example, based on the information collected from the lower proxy 230, and performs learning that can perform processing to eliminate problems generated in the routing. For example, the intermediate proxy 220 buffers the message when there is a message with the same To address whose destination address is unknown, and releases part of the buffered message to the network when the buffered amount exceeds a threshold. As a result, in the case of returning again, buffering is performed until it arrives, or buffering is performed for a certain period of time. Then, when messages start to arrive, the messages with the same To address in the buffered messages start to be sent.
[0081] The upper agent 210 performs learning for detecting a problem in the network, for example, based on information collected from the intermediate agent 220. Then, when a problem in the network is detected using the learning result, the upper agent 210 notifies the operator of the network of the detection result, etc., or outputs an instruction to change the network configuration.
[0082] For example, when the autonomous distributed system 200 has the purpose of implementing control related to automatic driving, each of the plurality of lower-level agents 230 collects information from IoT devices 400 mounted on vehicles located in the region, IoT devices 400 mounted on traffic lights, etc., and IoT devices 400 installed on roads, etc., for each region divided into a plurality of regions. As a specific example, the lower-level agent 230 collects vehicle position information, images captured by a camera of a vehicle or a camera on a street, vehicle detection information detected by a sensor on a road, human detection information, distance between vehicles detected by a sensor on a vehicle, etc., weather information in the region, navigation information of the vehicle, and driving speed of the vehicle, etc. In this case, the lower-level agent 230 is responsible for a certain region, and when a certain vehicle enters the region, it collects information from the IoT device 400 mounted on the vehicle, and when the vehicle leaves the region, the lower-level agent 230 in charge of the region the vehicle enters takes over the collection of information.
[0083] The lower agent 230 shares information with other lower agents 230, and performs detection of dangers such as the inter-vehicle distance being shorter than a threshold, a collision between a vehicle and a person, etc. When a danger is detected, the lower agent 230 transmits danger detection information to the target vehicle. The vehicle issues a warning to the driver or stops driving according to the received danger detection information.
[0084] The intermediate agent 220, for example, instructs the lower agent 230 on what information to communicate. The intermediate agent 220, for example, analyzes (learns) important information (many people, many cars, etc.) based on information (day of the week, time, etc.) collected from the lower agent 230 in the past, and can determine the information with high priority to be collected. Then, the intermediate agent 220 issues instructions to the lower agent 230 so that the lower agent 230 can collect high-priority information in each period. The intermediate agent 220 learns by cooperating with other intermediate agents 220 and can determine high-priority information with high accuracy. As a specific example, as a result of learning, the intermediate agent 220 instructs in the morning on weekdays so that the priority of the car detection results can be increased to collect information, and instructs in the daytime on weekends so that the priority of the person detection results can be increased to collect information.
[0085] The upper agent 210 performs a wider range of analysis, for example, and performs a forecast (at a set time) after several steps based on traffic information (traffic volume, time, day of the week, events, weather) of the responsible area (city, county, country), and provides information to the intermediate agent 220.
[0086] Figure 7 Another example of the autonomous distributed system 200 is schematically shown. The autonomous distributed system 200 may further include a switching agent 250. The switching agent 250 may control the amount of information flowing in the communication network 100. For example, when the routing of messages is controlled by the upper agent 210, the middle agent 220, and the lower agent 230, the switching agent 250 performs control of the communication flow.
[0087] Figure 8 An example of the hardware structure of the computer 1200 that implements the upper agent 210, the intermediate agent 220, or the lower agent 230 is schematically shown. The program installed in the computer 1200 enables the computer 1200 to function as one or more "parts" of the device involved in this embodiment, or enables the computer 1200 to perform operations associated with the device involved in this embodiment or the one or more "parts", and / or enables the computer 1200 to perform the process involved in this embodiment or the stage of the process. In order to make the computer 1200 perform specific operations associated with some or all of the blocks in the flowcharts and block diagrams described in this specification, such a program can be executed by the CPU 1212.
[0088] The computer 1200 of this embodiment includes a CPU 1212, a RAM 1214, and a graphics controller 1216, which are connected to each other through a main controller 1210. The computer 1200 also includes a communication interface 1222, a storage device 1224, a DVD drive 1226, and an input / output unit such as an IC card drive, which are connected to the main controller 1210 via an input / output controller 1220. The DVD drive 1226 can be a DVD-ROM drive and a DVD-RAM drive, etc. The storage device 1224 can be a hard disk drive and a solid state drive, etc. The computer 1200 also includes a ROM 1230 and a conventional input / output unit such as a keyboard, which are connected to the input / output controller 1220 via an input / output chip 1240.
[0089] The CPU 1212 controls each unit by operating according to the program stored in the ROM 1230 and the RAM 1214. The graphic controller 1216 acquires image data generated by the CPU 1212 in a frame buffer or the like provided in the RAM 1214 or in itself, and displays the image data on the display device 1218.
[0090] The communication interface 1222 communicates with other electronic devices via a network. The storage device 1224 stores programs and data used by the CPU 1212 in the computer 1200. The DVD drive 1226 reads programs or data from the DVD-ROM 1227 and provides them to the storage device 1224. The IC card drive reads programs and data from an IC card and / or writes programs and data to an IC card.
[0091] The ROM 1230 stores therein a boot program or the like executed by the computer 1200 when activated and / or a program depending on the hardware of the computer 1200. The input-output chip 1240 may also connect various input-output units to the input-output controller 1220 via a USB port, a parallel port, a serial port, a keyboard port, a mouse port, and the like.
[0092] The program is provided by a computer-readable storage medium such as DVD-ROM 1227 or an IC card. The program is read from the computer-readable storage medium and installed in the storage device 1224, RAM 1214 or ROM 1230, examples of which are computer-readable storage media, and executed by the CPU 1212. The information processing described in these programs is read into the computer 1200, resulting in cooperation between the program and the various types of hardware resources described above. The device or method can be configured by implementing the operation or processing of information according to the use of the computer 1200.
[0093] For example, in the case of performing communication between the computer 1200 and an external device, the CPU 1212 can execute a communication program loaded into the RAM 1214, and instruct the communication interface 1222 to perform communication processing based on the processing described in the communication program. Under the control of the CPU 1212, the communication interface 1222 reads transmission data stored in a transmission buffer area provided in a recording medium such as the RAM 1214, the storage device 1224, the DVD-ROM 1227, or an IC card, and transmits the read transmission data to the network, or writes reception data received from the network to a reception buffer area provided on the recording medium, etc.
[0094] Furthermore, the CPU 1212 can read all or a necessary part of a file or database stored in an external recording medium such as the storage device 1224, the DVD drive 1226 (DVD-ROM 1227), or an IC card to the RAM 1214, and perform various types of processing on the data on the RAM 1214. Then, the CPU 1212 can write the processed data back to the external recording medium.
[0095] Various types of information such as various types of programs, data, tables, and databases can be stored in the recording medium and receive information processing. CPU1212 can perform various types of processing described in various places of this disclosure, including various types of operations specified by the instruction sequence of the program, information processing, conditional judgment, conditional branching, unconditional branching, information retrieval / replacement, etc., for the data read from RAM1214, and write the results back to RAM1214. In addition, CPU1212 can retrieve information in files, databases, etc. in the recording medium. For example, in the case where multiple entries each having an attribute value of a first attribute associated with an attribute value of a second attribute are stored in the recording medium, CPU1212 can retrieve an entry that is consistent with the condition specifying the attribute value of the first attribute from the multiple entries, and read the attribute value of the second attribute stored in the entry, thereby obtaining the attribute value of the second attribute associated with the first attribute that meets the predetermined condition.
[0096] The program or software module described above may be stored in a computer-readable storage medium on or near the computer 1200. In addition, a recording medium such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet can be used as a computer-readable storage medium, thereby providing the program to the computer 1200 via the network.
[0097] The blocks in the flowcharts and block diagrams in this embodiment may represent the stages of the process of performing an operation or the "parts" of a device having the function of performing an operation. Specific stages and "parts" may be implemented by dedicated circuits, programmable circuits supplied together with computer-readable instructions stored on a computer-readable storage medium, and / or processors supplied together with computer-readable instructions stored on a computer-readable storage medium. Dedicated circuits may include digital and / or analog hardware circuits, and may also include integrated circuits (ICs) and / or discrete circuits. Programmable circuits may include, for example, field programmable gate arrays (FPGAs) and programmable logic arrays (PLAs), including reconfigurable hardware circuits such as logical products, logical sums, exclusive ors, and-nots, or-nots, and other logical operations, triggers, registers, and storage elements.
[0098] A computer-readable storage medium may include any tangible device capable of storing instructions to be executed by an appropriate device, and as a result, a computer-readable storage medium having instructions stored therein has a product containing instructions executable in order to make a means for performing the operations specified by the flowchart or block diagram. Examples of computer-readable storage media may include electronic storage media, magnetic storage media, optical storage media, electromagnetic storage media, semiconductor storage media, etc. More specific examples of computer-readable storage media may include: floppy disk (registered trademark), magnetic disk, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), electrically erasable programmable read-only memory (EEPROM), static random access memory (SRAM), compact disk read-only memory (CD-ROM), digital versatile disk (DVD), Blu-ray (registered trademark) disk, memory stick, integrated circuit card, etc.
[0099] Computer readable instructions may include assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state setting data, or any of source code or object code described by any combination of one or more programming languages, including object-oriented programming languages such as Smalltalk (registered trademark), JAVA (registered trademark), C++, etc., and existing procedural programming languages such as the "C" programming language or similar programming languages.
[0100] The computer readable instructions can be provided to a processor or programmable circuit of a general-purpose computer, a special-purpose computer or other programmable data processing device locally or via a wide area network (WAN) such as a local area network (LAN), the Internet, etc., so that the processor or programmable circuit of the general-purpose computer, special-purpose computer or other programmable data processing device executes the computer readable instructions in order to generate means for performing the operations specified in the flowchart or block diagram. Examples of processors include computer processors, processing units, microprocessors, digital signal processors, controllers, microcontrollers, etc.
[0101] The present invention has been described above using the embodiments, but the technical scope of the present invention is not limited to the scope described in the above embodiments. It is obvious to those skilled in the art that various changes or improvements can be made to the above embodiments. According to the description of the claims, the method of making such changes or improvements can also be included in the technical scope of the present invention.
[0102] The execution order of each process such as actions, processes, steps and stages in the apparatus, system, program and method shown in the claims, specifications and drawings is not specifically indicated as "earlier", "before", etc. In addition, it should be noted that as long as the output of the previous process is not used in the subsequent process, it can be implemented in any order. Even if the action flow in the claims, specifications and drawings is described using "first," "next," etc. for the sake of convenience, it does not mean that it must be implemented in this order.
[0103] Description of Reference Numerals
[0104] 100: network; 110: upper computing; 120: intermediate computing; 130: lower computing; 200: autonomous distributed system; 210: upper agent; 211: storage; 212: registration; 213: information collection; 214: learning execution; 215: processing execution; 216: information transmission; 220: intermediate agent; 221: storage; 222: registration; 223: information collection; 224: learning execution; 225: processing execution; 226: information transmission; 230: lower agent; 231: storage; 232: registration; 233: information collection; 234: Learning execution unit; 235: processing execution unit; 236: information sending unit; 250: switching agent; 300: cloud network; 310: cloud computing; 320: fog computing; 330: edge computing; 400: IoT device; 1200: computer; 1210: main controller; 1212: CPU; 1214: RAM; 1216: graphics controller; 1218: display device; 1220: input and output controller; 1222: communication interface; 1224: storage device; 1226: DVD drive; 1227: DVD-ROM; 1230: ROM; 1240: input and output chip.
Claims
1. A system, characterized in that: have: a first-level proxy configured at a first level of the network and a plurality of second-level proxies at a second level lower than the first level, Each of the plurality of second-level agents uses the collected information to perform learning in cooperation with other second-level agents, and transmits information generated using the learning result to the first-level agent. The first level agent performs learning using information received from the plurality of second level agents.
2. The system according to claim 1, wherein: The system further comprises a plurality of third-level agents at a third level lower than the second level, Each of the plurality of third-level agents uses the collected information to perform learning in cooperation with other third-level agents, and transmits information generated using the learning result to at least any one of the plurality of second-level agents.
3. The system according to claim 2, wherein: Each of the plurality of third-level agents has: The third information collection department collects information; a third learning execution unit that uses the information collected by the third information collection unit to perform learning in collaboration with other third-level agents; as well as a third information transmitting unit configured to transmit information generated using the learning result of the third learning executing unit to at least one of the plurality of second-level agents; Each of the plurality of second-level agents has: A second information collection unit collects information from a plurality of the third-level agents; a second learning execution unit that uses the information collected by the second information collection unit to perform learning in cooperation with other plurality of second-level agents; as well as a second information sending unit, which sends information generated using the learning result of the second learning execution unit to the first hierarchical agent; The first-tier agent has: a first information collecting unit, collecting information from a plurality of the second-level agents; as well as The first learning execution unit executes learning using the information collected by the first information collection unit.
4. The system according to claim 3, wherein: The first-level proxy is configured at the first NW level of the hierarchical network, The plurality of second-level agents are arranged at a second NW level lower than the first NW level of the hierarchical network, The plurality of third-level agents are arranged at a third NW level lower than the second NW level of the hierarchical network.
5. The system according to claim 4, wherein: The hierarchical network is a cloud network. The first NW level consists of cloud computing. The second NW layer consists of multiple fog computing components. The third NW level is composed of multiple edge calculations.
6. The system according to claim 4, wherein: The hierarchical network is a cloud network. The first NW level consists of cloud computing. The second NW layer consists of multiple fog computing components. The third NW level is composed of multiple edge computing. The third information collection unit collects information from multiple IoT devices through mobile communications, The third information sending unit uses the learning result of the third learning execution unit to send information selected from the information collected by the third information collecting unit or information generated using the learning result to at least any one of the plurality of second-level agents, and the second information collecting unit collects the information sent by the third information sending unit, The second information transmitting unit transmits information selected from the information collected by the second information collecting unit or information generated using the learning result to the first hierarchical agent using the learning result of the second learning executing unit.
Citation Information
Patent Citations
User-adaptive action determination device and action determination method
JP2004227208A
IOT-oriented fog computing architecture
CN105610944A
Fog computing environment-based vehicle prediction and diagnosis method
CN111709542A
Cloud, fog and edge end cooperative bearing state monitoring and management method and system
CN113465920A
