Machine Learning Method, System, Base Station and Medium Based on B-M2M

Abstract

The present disclosure provides a machine learning method, system, base station, and computer-readable storage medium based on B-M2M. Among them, the method includes: constructing a broadcast machine-to-machine B-M2M network architecture, where the B-M2M network architecture includes a B-M2M channel through which each device node can perform broadcast communication; and sending machine learning instructions to each device node based on the B-M2M channel, so that each device node cooperatively executes tasks according to the machine learning instructions based on the B-M2M channel to obtain the execution results of each device node. By constructing the B-M2M network architecture and using the B-M2M channel to realize broadcast communication between devices, the embodiments of the present disclosure can at least solve the problems that current industrial field robots do not have strong learning capabilities and it is difficult to achieve multi-robot data sharing, and effectively improve the machine learning efficiency.

Description

Technical Field

[0001] The present disclosure relates to the field of communication technologies, and in particular, to a machine learning method based on B-M2M, a machine learning system based on B-M2M, a base station, and a computer-readable storage medium. Background Art

[0002] Industrial Internet is an industrial and application ecosystem formed by the all-round deep integration of the Internet and new generation information technologies with industrial systems. It is a key comprehensive information infrastructure for the intelligent development of industry and is closely connected with intelligent manufacturing. With the development of the Industrial Internet and artificial intelligence, the automated operations in industrial sites have changed from traditional automatic control to various robots with machine learning, replacing manual operations in traditional complex situations, such as handling, palletizing, welding, spraying, assembly, cutting and grinding, etc., in a more flexible way. Currently, the working mode of robots mainly adopts the on-site teaching mode, which requires experienced operators and consumes a large amount of time, and the path may not be optimal. When the workpieces and the working environment are in dynamic changes, the on-site teaching mode cannot be completed, and robots need to have strong machine learning capabilities.

[0003] Currently, the machine learning of robots adopts a single closed mode and cannot share machine learning data with other similar robots. It is similar to students in a class studying alone without being able to communicate and share learning experiences and achievements with each other, which affects the learning efficiency. For existing network communications, taking 5G (5th Generation Mobile Communication Technology) network communication as an example, it is mainly oriented to point-to-point communication. If it is necessary to enter the network communication broadcast mode, it usually needs to add a new air interface at the network layer. However, for the node broadcast of a large number of robots in industrial sites, it will lead to problems such as low efficiency and high cost. Summary of the Invention

[0004] The present disclosure provides a machine learning method, system, base station, and computer-readable storage medium based on B-M2M to at least solve the problems such as low learning efficiency caused by the lack of strong learning capabilities of robots in industrial sites and the difficulty in sharing data among multiple robots.

[0005] According to one aspect of the present disclosure, there is provided a machine learning method based on B-M2M, including:

[0006] Constructing a broadcast machine-to-machine B-M2M network architecture, where the B-M2M network architecture includes a B-M2M channel through which each device node can perform broadcast communication; and,

[0007] Send machine learning instructions to each device node based on the B-M2M channel, so that each device node collaboratively executes tasks according to the machine learning instructions based on the B-M2M channel, and obtains the execution results of each device node.

[0008] In one implementation, the device node includes a mechanical device, a mechanical controller, a data acquisition device, and an output execution mechanism.

[0009] In one implementation, after sending the machine learning instructions to each device node based on the B-M2M channel, it further includes:

[0010] Determine whether the execution results of each device node are all qualified;

[0011] If the execution result of a certain device node is unqualified, evaluate the execution results of each device node to obtain the evaluation results of each device node;

[0012] Add the evaluation results to the machine learning instructions, and return to the step of sending the machine learning instructions to each device node based on the B-M2M channel, so that each device node collaboratively executes the machine learning instructions according to the evaluation results based on the B-M2M channel until the execution results of all device nodes are qualified.

[0013] In one implementation, the method further includes:

[0014] Pre-store a number of machine learning methods, and the number of machine learning methods includes model-based machine learning methods and model-free machine learning methods;

[0015] Select the corresponding machine learning method from the number of machine learning methods;

[0016] The evaluation of the execution results of each device node includes:

[0017] Evaluate the execution results of each device node based on the selected machine learning method.

[0018] In one implementation, the evaluation results include a first evaluation result and a second evaluation result, where the second evaluation result carries improvement data, the first evaluation result indicates that the corresponding device node continues to execute the task based on the current machine learning instruction, and the second evaluation result indicates that the corresponding device node continues to execute the task based on the improvement data.

[0019] According to another aspect of the present disclosure, a B-M2M-based machine learning system is provided, including:

[0020] A building block configured to build a broadcast machine-to-machine (B-M2M) network architecture, the B-M2M network architecture including a B-M2M channel through which broadcast communication can be carried out between various device nodes in a preset area; and,

[0021] A sending module configured to send machine learning instructions to each device node based on the B-M2M channel, so that each device node cooperatively executes tasks according to the machine learning instructions based on the B-M2M channel to obtain execution results of each device node.

[0022] In one implementation, the system further includes:

[0023] A judging module configured to judge whether the execution results of all device nodes are qualified after the sending module sends machine learning instructions to each device node;

[0024] An evaluating module configured to, when the judging module judges that there is an unqualified execution result of a certain device node, evaluate the execution results of all device nodes to obtain evaluation results of all device nodes;

[0025] An adding and returning module configured to add the evaluation results to the machine learning instructions and return the sending module to send machine learning instructions to each device node based on the B-M2M channel, so that each device node cooperatively executes the machine learning instructions according to the evaluation results based on the B-M2M channel until the execution results of all device nodes are qualified.

[0026] In one implementation, the system further includes:

[0027] A pre-storing module configured to pre-store a plurality of machine learning methods, the plurality of machine learning methods including model-based machine learning methods and model-free machine learning methods;

[0028] A selecting module configured to select a corresponding machine learning method from the plurality of machine learning methods;

[0029] The evaluating module is specifically configured to evaluate the execution results of all device nodes based on the selected machine learning method.

[0030] According to another aspect of the present disclosure, there is provided a base station including a memory and a processor, wherein a computer program is stored in the memory, and when the processor runs the computer program stored in the memory, the processor executes the above-mentioned B-M2M-based machine learning method.

[0031] According to another aspect of the present disclosure, there is provided a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the processor executes the above-described B-M2M-based machine learning method.

[0032] The technical solutions provided by the present disclosure may include the following beneficial effects:

[0033] The B-M2M-based machine learning method provided by the present disclosure, by constructing a B-M2M network architecture and utilizing the broadcast communication mode of the B-M2M channel, can support efficient information sharing between devices without adding a new air interface at the network layer. The task execution process of obtaining machine learning instructions based on the B-M2M channel and sharing machine learning instructions with other device nodes can at least solve the problems that current industrial field robots do not have strong learning capabilities and it is difficult to achieve multi-robot data sharing, effectively improving the machine learning efficiency.

[0034] Other features and advantages of the present disclosure will be described in the following description, and in part will be obvious from the description, or will be understood by implementing the present disclosure. The objectives and other advantages of the present disclosure can be achieved and obtained by the structures specifically pointed out in the description, claims and drawings. Description of the Drawings

[0035] The drawings are used to provide a further understanding of the technical solutions of the present disclosure, and constitute a part of the description. They are used together with the embodiments of the present disclosure to explain the technical solutions of the present disclosure, and do not constitute a limitation to the technical solutions of the present disclosure.

[0036] Figure 1 It is a schematic flow chart of a B-M2M-based machine learning method provided by an embodiment of the present disclosure;

[0037] Figure 2 It is a schematic scenario diagram of the B-M2M network architecture in the present disclosure;

[0038] Figure 3 It is a schematic flow chart of another B-M2M-based machine learning method provided by an embodiment of the present disclosure;

[0039] Figure 4 It is a schematic structural diagram of a B-M2M-based machine learning system provided by an embodiment of the present disclosure;

[0040] Figure 5 It is a schematic structural diagram of a base station provided by an embodiment of the present disclosure. Detailed Embodiments

[0041] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure clearer, the following provides a detailed description of the specific embodiments of the present disclosure with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for the purpose of illustrating and explaining the present disclosure, and are not intended to limit the present disclosure.

[0042] It should be noted that the terms "first", "second", etc. in the description and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence; moreover, in the case of no conflict, the embodiments in the present disclosure and the features in the embodiments can be arbitrarily combined with each other.

[0043] In subsequent descriptions, the use of suffixes such as "module", "component", or "unit" to represent elements is only for the convenience of describing the present disclosure, and they have no specific meaning in themselves. Therefore, "module", "component", or "unit" can be used interchangeably.

[0044] Industrial Internet and machine learning require high-performance wireless communication networks and powerful data processing capabilities as support. The fifth-generation mobile communication (5G) features low latency, high reliability, and large capacity. 5G has significantly improved compared to 4G in terms of peak rate, latency, user experience rate, and the number of simultaneously supported connections. The integrated and innovative development of 5G and the Industrial Internet will drive the transformation of manufacturing from single-point and local information technology applications to digital, networked, and intelligent applications, and also open up a broader market space for 5G.

[0045] Machine learning is a multi-disciplinary field that involves multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. Machine learning simulates human learning behaviors to acquire new knowledge or skills. Machine learning is one of the most intelligent and cutting-edge research fields in artificial intelligence. According to different learning methods, machine learning can be divided into: (1) Supervised learning (learning with a teacher): There is a teacher signal in the input data. Using probability functions, algebraic functions, or artificial neural networks as the basis function models and adopting iterative calculation methods, the learning result is a function; (2) Unsupervised learning (learning without a teacher): There is no teacher signal in the input data. Using clustering methods, the learning result is a category. Typical unsupervised learning includes discovery learning, clustering, competitive learning, etc.; (3) Reinforcement learning: A learning method that uses environmental feedback (reward / punishment signals) as input and is guided by statistical and dynamic programming techniques.

[0046] Currently, machine learning in industrial sites mainly faces the following problems: Machine learning adopts a single closed mode and cannot share data with other similar machine learning systems. It is like students in a class studying alone without being able to communicate and share their learning experiences and achievements with each other, which affects the learning efficiency. Existing 5G mainly focuses on point-to-point communication, and the broadcast communication efficiency is relatively low. Process evaluation and control algorithms require a large amount of computing and processing (such as image recognition, decision-making algorithms, optimization algorithms). If they are placed on the device side, it will increase the complexity and cost of the device side, resulting in a sharp increase in the cost of the entire system.

[0047] Therefore, according to the characteristics of the 5G network, the embodiments of the present disclosure construct a Broadcast Machine-to-Machine (B-M2M) architecture based on the 5G network to achieve an efficient broadcast communication mode and support efficient information sharing between devices. Based on the efficient broadcast communication mode, real-time information sharing and mutual learning between devices can be realized, which can significantly improve the efficiency of machine learning. By making full use of the powerful data processing and storage capabilities of the Mobile Edge Computing (MEC) of 5G, machine learning is placed in the MEC, thereby endowing the device side with powerful machine learning capabilities and effectively reducing the cost of the device side and increasing flexibility. The embodiments of the present disclosure construct a new machine broadcast communication (B-M2M) mode, reconstruct the machine learning strategy under this mode, and then apply this technology to various typical applications such as machine learning in industrial sites, thereby building a technical ecosystem of B-M2M, enriching the development space of the 5G network, which has positive significance.

[0048] Please refer to Figure 1 , Figure 1 FIG. Figure 1 is a schematic flowchart of a machine learning method based on B-M2M provided by the embodiments of the present disclosure. As

[0049] shown, it is applied to a base station, and the method includes step S101 and step S102.

[0050] In step S101, a broadcast machine-to-machine (B-M2M) network architecture is constructed, and the B-M2M network architecture includes a B-M2M channel through which broadcast communication can be carried out between each device node.

[0051] In this embodiment, the B-M2M network architecture is a 5G-based B-M2M network architecture. By using the licensed frequency band of 5G, within the coverage area of the industrial field base station, dedicated frequency bands are dynamically divided, and the common broadcast time slot channel (B-M2M channel) is deployed in a time-division manner. All device nodes within the network (i.e., within the device area covered by the base station network) have the ability to receive all broadcast time slots. Each device node can dynamically select an idle time slot in the B-M2M channel to send broadcast information, thus realizing the broadcast sending and receiving of all device nodes.

[0052] In addition to the ability of each device node to perform broadcast communication with each other, broadcast communication can also be carried out between the base station and each device node through this B-M2M channel. Further, the base station can configure dedicated control time slots. Based on these control time slots, the base station can dynamically adjust the frequency band width and the number of time slots of the resource pool according to the real-time broadcast intensity to ensure that the broadcast transmission delay of each device node meets the quality requirements of the production site.

[0053] As Figure 2 shown, the B-M2M network architecture may include: device nodes, a common broadcast time slot channel, and a B-M2M management unit (not shown) deployed on the base station side. Among them, device nodes: have the function of wireless broadcast information sending and receiving, and are installed at each core part of industrial production equipment. All device nodes have the function of receiving all time slots of the common broadcast channel resource pool, such as robots; common broadcast time slot channel: within the coverage area of the base station, it is a common broadcast channel resource pool with continuous frequency bands and time slots managed by the B-M2M management unit in the base station. The frequency band width and the number of time slots of the resource pool are dynamically adjusted by the B-M2M management unit according to the real-time broadcast intensity to ensure that the broadcast transmission delay of each device node meets the quality requirements of the production site; B-M2M management unit: deployed in the base station. Specifically, the B-M2M management unit is divided into a B-M2M broadcast sending and receiving module and a B-M2M management and control system, and the two are respectively deployed in the access network of the base station and the mobile edge computing MEC of the base station. For example, the B-M2M broadcast sending and receiving module (hereinafter referred to as the B-M2M module) is deployed in the access network (5G NG-RAN) of the 5G base station, and has the functions of broadcast management information, confirmation information, and status information, system management, and the function of receiving all time slots of the common broadcast channel resource pool. The mobile edge computing platform of the base station deploys the B-M2M management and control system and the operation of the production application system.

[0054] In addition, a cloud computing platform can also be deployed in the 5G core network (5GC) to complete data offloading. Specifically, the MEC of the base station is connected to the cloud platform through the mobile core network. The parameters, management, and control algorithms in the MEC are uniformly offloaded from the cloud platform. The MEC is responsible for operation and reporting of the operation status. This mode ensures the consistency of the entire system and the rapid upgrade of functions. Various production application systems are published in the application store of the cloud platform and are offloaded to the MEC for deployment and operation according to the needs of the production site.

[0055] In one implementation, the device node includes a mechanical device, a mechanical controller, a data acquisition device, and an output execution mechanism.

[0056] Specifically, the mechanical device: has an automatic control function and can complete corresponding actions according to input instructions, such as automatic plug-in machines, automatic robotic arms, handling robots, welding robots, etc. in belt assembly lines, assembly lines, plug-in lines, and assembly lines. The mechanical device is controlled by the output signal of the mechanical controller and completes specified actions under the control of the mechanical controller; the mechanical controller: after receiving a standard instruction data packet, processes the instruction packet, and then converts it into a control signal format and level that meets the input requirements of the mechanical device, and sends it to the mechanical device, enabling the mechanical device to complete certain actions or operation tasks, such as controlling the movement position, posture, trajectory, operation sequence, and action duration of an industrial robotic arm in the working space according to the input control signal. The mechanical controller is connected to the B-M2M device node and receives the control signal data packet through the B-M2M device node; the data acquisition device: various types of sensors, image acquisition and recognition devices are used to collect various data in the industrial field, and after passing through the on-site processing module, the collected data is converted into a B-M2M standard broadcast data packet and broadcast and sent through the B-M2M device node; the output execution mechanism: receives the control signal received by the B-M2M device terminal, converts the control signal into a corresponding drive signal, and uses an electric, liquid, gas, or other energy device to convert it into a mechanical action (such as angular displacement or linear displacement to control valves and switches), or changes process parameters through an actuator (such as increasing the temperature of a container or extending the cooling time).

[0057] In this embodiment, the device nodes in the industrial field include the above four structures. The mechanical controller, data acquisition device, and other four structures deployed in the basic environment of the industrial field jointly implement the control function of the device node. It can be understood that each device node in this embodiment is a device node with the same function.

[0058] In step S102, machine learning instructions are sent to each device node based on the B-M2M channel, so that each device node collaboratively executes tasks according to the machine learning instructions based on the B-M2M channel, and obtains the execution results of each device node.

[0059] In this embodiment, a machine learning control algorithm is deployed in the MEC of the base station to avoid problems such as low machine learning efficiency and high costs caused by the robot itself having weak machine learning capabilities or low computing power. The base station uses the B-M2M channel to send machine learning instructions to each device node to guide each device node to perform machine learning. At the same time, each device node performs broadcast data sharing based on the B-M2M channel to improve the machine learning efficiency between device nodes and better execute tasks.

[0060] Furthermore, the base station receives the working status data broadcast by each device node in all time slots through the B-M2M module (i.e., the B-M2M broadcast sending and receiving module) of the B-M2M architecture. Furthermore (equivalent to the master being able to understand the working conditions of all apprentices at the same time), the B-M2M architecture enables the B-M2M module of the base station to broadcast and send data in multiple time slots (equivalent to the master being able to guide the working results of multiple apprentices at the same time).

[0061] Specifically, the control algorithm in the MEC sends instructions to the devices in the industrial field. The instructions are sent by selecting idle time slots through the B-M2M module of the base station. After the B-M2M modules of each device node in the industrial field receive the broadcast instruction data packet, they send the instruction data packet to its mechanical controller. The mechanical controller converts it into control execution signals for the mechanical equipment and the output execution mechanism, and the local execution unit controls and combines different parameters to control the mechanical equipment to complete this action. Taking the grinding of the welding joint of the carriage by the robotic arm as an example, the control algorithm in the MEC issues an instruction to grind the welding joint and attaches the initial working parameters. Through B-M2M broadcast, after the device node receives the instruction and the initial working parameters, it combines the eigenvalue identified by the local on-site working part video to generate modified parameters, such as the range of the up and down movement of the grinding head of the robotic arm and the forward position, generates a control signal that meets the command word format of the mechanical equipment and the output execution mechanism, and sends it to the mechanical equipment and the output execution mechanism to start the corresponding action. After the corresponding actions of the mechanical equipment and the output execution mechanism are completed, the data acquisition device in the basic environment of the industrial field performs image acquisition and data acquisition on the site after the action is executed, and the acquired results can be broadcast through the B-M2M module. Other device nodes and the MEC can receive these data, so that the MEC can control them and the mutual and collaborative learning between each device node can improve the learning efficiency of processing the next action.

[0062] Compared with the existing technology where each device node performs machine learning independently, in this embodiment, by constructing a B-M2M architecture and implementing broadcast communication between the base station and devices, and between devices based on the B-M2M channels in the B-M2M architecture, data broadcast communication can be achieved without adding a new air interface at the network layer. Moreover, leveraging the strong computing power and storage and processing capabilities of the MEC in the base station to guide the machine learning of each device node and the mutual learning between device nodes based on the broadcast mode effectively solves the problems in the prior art such as limited computing resources of robots, lack of strong learning ability, and low learning efficiency.

[0063] Please refer to Figure 3 , Figure 3 A machine learning method based on B-M2M provided in another embodiment of the present disclosure. Based on the previous embodiment, this embodiment utilizes the computing power of the MEC in the base station to evaluate the execution results fed back by each device node, thereby endowing each device node with strong learning ability and improving the efficiency of machine learning tasks. Specifically, after sending machine learning instructions to each device node based on the B-M2M channel (step S102), steps S301 - S303 are further included.

[0064] The machine learning architecture based on B-M2M in this embodiment is based on the reinforcement learning architecture and integrates the B-M2M network. Through the interaction between B-M2M, mechanical equipment, output actuators, and data acquisition devices, feedback signals can be given to the mechanical equipment and output actuators at any time. The mechanical controller, data acquisition devices, and B-M2M modules deployed in the industrial field basic environment implement the control function of the devices. The mechanical equipment and output actuators deployed in the industrial field basic environment receive the signals output by the mechanical controller and complete the actions. The B-M2M module and the machine learning algorithm in the mobile edge computing (MEC) deployed in the 5G network base station complete the evaluation of the actions and the generation of reward and punishment signals, and broadcast them through the B-M2M module of the base station.

[0065] In step S301, it is judged whether the execution results of all device nodes are qualified. If the execution result of a certain device node is unqualified, step S302 is executed; otherwise, the process ends.

[0066] Taking the grinding of the welding joint of the carriage by the robotic arm as an example, it is judged whether the grinding of the welding joint part meets the requirements of indicators such as surface finish, thickness, flatness, and height difference. If the execution result meets the indicator requirements, it indicates that the execution result of the device node is qualified, and each device node completes the task, and the system ends the process; otherwise, each device node enters step S302 to continue to execute the task.

[0067] In step S302, the execution results of each device node are evaluated to obtain the evaluation results of each device node.

[0068] Specifically, each device node broadcasts its execution result through the B-M2M module. After the B-M2M module of the base station receives the result data, it transmits it to the control algorithm in the MEC for processing such as image recognition and data acquisition. Then, the processing result of the data is comprehensively evaluated, and the reward and punishment information is generated according to the evaluation result. Finally, it is broadcast and sent through the B-M2M module of the base station. Taking the robotic arm grinding the welding joint of the carriage as an example, the grinding head of the robotic arm performs the grinding action, and the image acquisition device on-site captures the image of the grinding part. At the same time, the high-precision laser measurement sensor measures the flatness, multi-point height difference, and shape of the grinding part, and broadcasts the captured image and the data measured by the laser measurement sensor through the B-M2M module of the device node. After the B-M2M module of the base station receives these data, it transmits them to the evaluation algorithm of the MEC for evaluation, and then gives the evaluation result.

[0069] In step S303, the evaluation result is added to the machine learning instruction, and the step of sending the machine learning instruction to each device node based on the B-M2M channel is returned, so that each device node collaboratively executes the machine learning instruction according to the evaluation result based on the B-M2M channel until the execution results of all device nodes are qualified.

[0070] Specifically, the evaluation result includes a first evaluation result and a second evaluation result, where the second evaluation result carries improvement data, the first evaluation result instructs the corresponding device node to continue to execute the task based on the current machine learning instruction, and the second evaluation result instructs the corresponding device node to continue to execute the task based on the improvement data.

[0071] In this embodiment, the first evaluation result is a reward evaluation result, indicating that the data processing parameters in the current task execution process are correct, and the second evaluation result is a punishment evaluation result, indicating that the data processing parameters in the current task execution process can still be optimized. Taking the robotic arm grinding the welding joint of the carriage as an example, after the mechanical control in the industrial field receives the reward signal, it prompts that under the recognition feature value of the on-site part image, the grinding part and angle are correct. At this time, the robotic arm continues to grind at this angle (as indicated in the machine learning instruction), continuously recognizes the image of the grinding part, obtains the feature value, and regularly broadcasts and sends it through the B-M2M module of the device node. The B-M2M module of the base station continuously receives and evaluates, and then feedbacks the evaluation result through the B-M2M module of the base station. If the punishment information (carrying improvement data) is obtained, the robotic arm adjusts the parameters (such as depth and angle) based on the improvement data and then repeats the operation until the grinding of this part meets the index requirements such as smoothness, thickness, flatness, and height difference.

[0072] Further, during the evaluation process, device nodes can also obtain evaluation information from each other, improving their own machine learning capabilities. Due to B-M2M broadcast mode communication, during the operation of a field device node, its working state parameters, characteristic values, operation modes, and corresponding reward and punishment results can also be received by other device nodes. These device nodes synchronously retain the intermediate data and result data of the above machine learning process as the learning data for the machine learning of this device node, which can significantly improve the learning ability of this device node. Taking the example of a robotic arm grinding the welding joint of a carriage, a certain device node may not have processed a certain shape of the grinding part. As long as other device nodes have encountered it, this device node can handle it well.

[0073] The machine learning of this embodiment updates the relevant parameters of the control system through the reporting, feedback, and mutual learning of B-M2M broadcasts, thereby adjusting the previous behavior. Through continuous adjustment and accumulation, it learns what kind of behavior to choose under what circumstances to obtain the best results.

[0074] Further, to meet the machine learning requirements of different device nodes in the industrial field, this embodiment flexibly deploys a variety of machine learning algorithms in the MEC to adapt to different industrial needs. The method further includes the following steps:

[0075] Pre-store a number of machine learning methods, where the number of machine learning methods includes model-based machine learning methods and model-free machine learning methods;

[0076] Select the corresponding machine learning method from the number of machine learning methods.

[0077] The evaluation of the execution results of each device node is specifically:

[0078] Evaluate the execution results of each device node based on the selected machine learning method.

[0079] Specifically, the base station can select the corresponding machine learning method according to the machine learning instruction and the current industrial field to evaluate the execution results of the device node.

[0080] Taking an industrial field such as a robotic arm grinding the welding joint of a carriage as an example, since the environment is relatively fixed, a model-based machine learning method (model-based) can be deployed, which can take the behavior sequence with the greatest benefit in a fixed environment. For an environment that is changing, a model-free machine learning method can be deployed to make the machine learning more versatile, such as the control of a transport robot.

[0081] Based on the same inventive concept, embodiments of the present disclosure correspondingly further provide a machine learning system based on B-M2M, as Figure 4 shown. The system includes a construction module 41 and a sending module 42, wherein,

[0082] The construction module 41 is configured to construct a broadcast machine-to-machine (B-M2M) network architecture, and the B-M2M network architecture includes a B-M2M channel through which each device node in a preset area can perform broadcast communication; and,

[0083] The sending module 42 is configured to send machine learning instructions to each device node based on the B-M2M channel, so that each device node cooperatively executes tasks according to the machine learning instructions based on the B-M2M channel, and obtains execution results of each device node.

[0084] In one implementation, the device node includes a mechanical device, a mechanical controller, a data acquisition device, and an output execution mechanism.

[0085] In one implementation, the system further includes:

[0086] A judgment module, which is configured to judge whether the execution results of each device node are all qualified after the sending module sends machine learning instructions to each device node;

[0087] An evaluation module, which is configured to evaluate the execution results of each device node to obtain evaluation results of each device node when the judgment module judges that there is an unqualified execution result of a certain device node;

[0088] An adding and returning module, which is configured to add the evaluation results to the machine learning instructions, and return the sending module to send machine learning instructions to each device node based on the B-M2M channel, so that each device node cooperatively executes the machine learning instructions according to the evaluation results based on the B-M2M channel until the execution results of all device nodes are qualified.

[0089] In one implementation, the system further includes:

[0090] A pre-storage module, which is configured to pre-store a plurality of machine learning methods, and the plurality of machine learning methods include a model-based machine learning method and a model-free machine learning method;

[0091] A selection module, which is configured to select a corresponding machine learning method from the plurality of machine learning methods;

[0092] The evaluation module is specifically configured to evaluate the execution results of each device node based on the selected machine learning method.

[0093] In one embodiment, the evaluation results include a first evaluation result and a second evaluation result, where the second evaluation result carries improvement data, the first evaluation result indicates that the corresponding device node continues to execute a task based on the current machine learning instruction, and the second evaluation result indicates that the corresponding device node continues to execute the task based on the improvement data.

[0094] Based on the same technical concept, an embodiment of the present disclosure correspondingly further provides a base station, as Figure 5 shown, including a memory 51 and a processor 52. A computer program is stored in the memory 51. When the processor 52 runs the computer program stored in the memory 51, the processor 52 executes the B-M2M-based machine learning method described above.

[0095] Based on the same technical concept, an embodiment of the present disclosure correspondingly further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the processor executes the B-M2M-based machine learning method.

[0096] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, and appropriate combinations thereof. In the hardware implementation, the division between the functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or one function or step may be executed by several physical components in cooperation. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassette, tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, a communication medium typically contains computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and may include any information delivery medium.

[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them; although the present disclosure has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A machine learning method based on B-M2M, characterized in that, applied to a base station, comprising: Constructing a broadcast machine-to-machine (B-M2M) network architecture, the B-M2M network architecture including a B-M2M channel through which each device node can perform broadcast communication; wherein, within the coverage area of the industrial field base station, a dedicated frequency band is dynamically divided, and the B-M2M channel is deployed in a time-division manner. All device nodes within the device area of the base station network coverage area have the ability to receive all broadcast time slots, and each device node can dynamically select an idle time slot in the B-M2M channel to send and receive broadcast information; and, Sending machine learning instructions to each device node based on the B-M2M channel, so that each device node collaboratively executes tasks according to the machine learning instructions based on the B-M2M channel, and obtains the execution results of each device node.

2. The method according to claim 1, characterized in that, the device node includes a mechanical device, a mechanical controller, a data acquisition device, and an output actuator.

3. The method according to claim 1, characterized in that, after sending machine learning instructions to each device node based on the B-M2M channel, further comprising: Judging whether the execution results of each device node are all qualified; If the execution result of a certain device node is unqualified, then evaluate the execution results of each device node to obtain the evaluation results of each device node; Add the evaluation results to the machine learning instructions, and return to the step of sending machine learning instructions to each device node based on the B-M2M channel, so that each device node collaboratively executes the machine learning instructions according to the evaluation results based on the B-M2M channel until the execution results of all device nodes are qualified.

4. The method according to claim 3, characterized in that, further comprising: Pre-storing a number of machine learning methods, the number of machine learning methods including model-based machine learning methods and model-free machine learning methods; Selecting a corresponding machine learning method from the number of machine learning methods; The evaluating the execution results of each device node includes: Evaluating the execution results of each device node based on the selected machine learning method.

5. The method according to claim 3 or 4, characterized in that, the evaluation results include a first evaluation result and a second evaluation result, wherein the second evaluation result carries improvement data, the first evaluation result instructs the corresponding device node to continue to execute the task based on the current machine learning instruction, and the second evaluation result instructs the corresponding device node to continue to execute the task based on the improvement data.

6. A machine learning system based on B-M2M, characterized in that, applied to a base station, comprising: A building block, which is configured to build a broadcast machine-to-machine B-M2M network architecture, and the B-M2M network architecture includes a B-M2M channel capable of performing broadcast communication between various device nodes in a preset area; wherein, within the coverage range of an industrial field base station, a dedicated frequency band is dynamically divided, and the B-M2M channel is deployed in a time-division manner. All device nodes in the device area within the coverage range of the base station network have the ability to receive all broadcast time slots, and each device node can dynamically select an idle time slot in the B-M2M channel to send and receive broadcast information; and, A sending module, which is configured to send machine learning instructions to each device node based on the B-M2M channel, so that each device node collaboratively executes tasks according to the machine learning instructions based on the B-M2M channel, and obtains the execution results of each device node.

7. The system according to claim 6, wherein, it further includes: A judging module, which is configured to judge whether the execution results of all device nodes are qualified after the sending module sends machine learning instructions to each device node; An evaluating module, which is configured to evaluate the execution results of each device node to obtain the evaluation results of each device node when the judging module judges that there is an unqualified execution result of a certain device node; An adding and returning module, which is configured to add the evaluation results to the machine learning instructions, and return the sending module to send machine learning instructions to each device node based on the B-M2M channel, so that each device node collaboratively executes the machine learning instructions according to the evaluation results based on the B-M2M channel until the execution results of all device nodes are qualified.

8. The system according to claim 7, wherein, it further includes: A pre-storing module, which is configured to pre-store a number of machine learning methods, and the number of machine learning methods includes a model-based machine learning method and a model-free machine learning method; A selecting module, which is configured to select a corresponding machine learning method from the number of machine learning methods; The evaluating module is specifically configured to evaluate the execution results of each device node based on the selected machine learning method.

9. A base station, wherein, it includes a memory and a processor, and a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the B-M2M-based machine learning method according to any one of claims 1 to 5.

10. A computer-readable storage medium, on which a computer program is stored, wherein, when the computer program is executed by a processor, the processor executes the B-M2M-based machine learning method according to any one of claims 1 to 5.

Images